# Agent Conf > Agent Conf is a 2026 conference in Warsaw, Poland for engineers and builders working with AI agents, agentic development, orchestration, and production AI workflows. Agent Conf takes place September 17-18, 2026 at Crowne Plaza Warsaw - The HUB. The event is hosted by Callstack and brings together 500+ engineers for talks, workshops, and hallway conversations about building software with agents. Key topics include vibe coding, agentic engineering fundamentals, orchestration and multi-agent systems, AI adoption, human checkpoints, debugging agent behavior, and moving agent workflows from experiments to production teams. ## Pages - [Homepage](https://agent.sh/): Main conference information, speakers, program topics, sponsor information, venue details, and hotel links. - [Agenda](https://agent.sh/agenda): Conference schedule for September 17-18, 2026. - [Workshop](https://agent.sh/workshop): Details for the September 16 hands-on software factory workshop. - [Tickets](https://agent.sh/tickets): Ticket checkout for Agent Conf 2026. - [Terms & Conditions](https://agent.sh/terms-conditions): Ticket purchase and conference participation terms. - [Privacy Policy](https://agent.sh/privacy-policy): Privacy information for attendees and website visitors. - [Code of Conduct](https://agent.sh/code-of-conduct): Participation standards for the conference. ## Event Facts - [Venue map](https://maps.google.com/?q=Crowne+Plaza+Warsaw+The+HUB): Crowne Plaza Warsaw - The HUB, Warsaw, Poland. - [Organizer](https://www.callstack.com/): Callstack, the company hosting Agent Conf. ## Agenda ### Thursday, September 17, 2026 - 09:15-09:25: Opening. - 09:25-09:50: Mike Grabowski and Michał Pierzchała — “Keynote”. - 09:50-10:15: Nader Dabit — “Navigating the Abundance Era of Software” (Orchestration & Multi-Agent Systems). - 10:15-10:40: Nick Miller — talk to be announced. - 10:40-11:10: Coffee Break. - 11:10-11:35: Jakub Kuzimski — “Stop Making LLMs Deterministic. Building a Spec-Driven Software Factory at Allegro.” (Agentic Engineering Fundamentals). - 11:35-12:00: David Mokos — “Your Biggest User Isn't Human Anymore: Designing DevTools for AI Agents” (Agentic Engineering Fundamentals). - 12:00-12:25: Dom Sipowicz — “Build an Agent with Eve: The Open Agent Framework” (Vibe Coding & Development). - 12:25-13:55: Lunch Break. - 13:55-14:20: Piotr Karwatka — “Agentic Coding at scale at Open Mercato” (Adoption & Case Studies). - 14:20-14:45: Anisha Malde and Kourtney Meiss — “Agents FC: Harness Your Agents Into a Winning Team” (Orchestration & Multi-Agent Systems). - 14:45-15:10: Zuzanna Skoczylas — “The AI Agent Iceberg” (Agentic Engineering Fundamentals). - 15:10-15:40: Coffee Break. - 15:40-16:05: Anjaneya Sastry Kappagantu — “Learning at Inference Time: TTT, TAO, δ-mem — and When to Spend a Gradient” (Agentic Engineering Fundamentals). - 16:05-16:30: Artur Morys-Magiera and Lech Kalinowski — “Quantization Impact On LLM Performance: A Case Study of Apex” (Adoption & Case Studies). - 16:30-16:55: David Budáč — “Has this happened before? - building a queryable memory of database logs with agents” (Agentic Engineering Fundamentals). - 16:55-17:20: Tejas Kumar — “The Critical Advantage With AI” (Agentic Engineering Fundamentals). - 17:20-19:05: Networking. ### Friday, September 18, 2026 - 09:10-09:15: Opening Day 2. - 09:15-09:40: Carl Ross — “No Token Budget: What Breaks When Agentic Software Factories Run on Kubernetes” (Orchestration & Multi-Agent Systems). - 09:40-10:05: Nishant Gupta — “The Agent Control Plane: Managing Thousands of AI Agents Without Losing Control” (Orchestration & Multi-Agent Systems). - 10:05-10:30: David Khourshid — “Goodbye Slop; Welcome Determinism” (Agentic Engineering Fundamentals). - 10:30-10:55: Kent C. Dodds — “I Built Your Personal Software Ecosystem” (Agentic Engineering Fundamentals). - 10:55-11:25: Coffee Break. - 11:25-11:50: Sven Großmann and Ivana Huckova — “When "Vibes" Aren't Enough: Observability for AI Agents” (Orchestration & Multi-Agent Systems). - 11:50-12:15: Sujith N S and Leonel Candido da Silva — “Agentic Monitoring in Production: How AI Agents Reduced Alert Investigation Time by 70%” (Adoption & Case Studies). - 12:15-12:40: Kitze — “From Vibe Coding to Vibe Engineering” (Vibe Coding & Development). - 12:40-14:10: Lunch Break. - 14:10-14:35: George Nikolaropoulos — “Building a Neobank in Under a Year: How Agentic Engineering Helped Us Ship Snappi” (Adoption & Case Studies). - 14:35-15:00: Ivan Charapanau — “How will we prompt AGI? A History of Harness Hacks.” (Orchestration & Multi-Agent Systems). - 15:00-15:25: Ruben Casas — “Beyond Components: The Future of AI UI” (Agentic Engineering Fundamentals). - 15:25-15:50: Levan Kvirkvelia — talk to be announced (Vibe Coding & Development). - 15:50-16:20: Coffee Break. - 16:20-16:45: Talk to be announced. - 16:45-17:20: Chris Trag, Margaret Pieczkowski, Bartek Pucek, Grzegorz Wolodko, and Carl Ross — “Panel Discussion: Future of Software Development.” - 17:20-17:30: Closing. - 18:00-00:00: After Party. ## Speakers ### Mike Grabowski - Company: Callstack - Talk: “Keynote” - Co-speakers: Michał Pierzchała - Links: [LinkedIn](https://www.linkedin.com/in/grabbou/) - Bio: Mike Grabowski is the CEO and co-founder of Callstack, a software engineering company that has spent the past decade helping enterprise teams build and scale products with React and React Native. He helped build the company from the ground up and led its technology as CTO for nearly ten years. In 2026, one year after Viking Global invested in Callstack, he became CEO. Callstack now has more than 230 people and exceeded PLN 150 million in revenue in 2025. Today, Callstack develops tools that help developers and AI agents work faster together and close the software development loop, from writing code to running and observing applications, testing, and validation. These include Agent Device, which allows agents to launch, operate, and test applications on devices and simulators, and Apex, a model specialized in React Native. Mike has invested in the community from the beginning. Callstack is a React Native core contributor and a founding member of the React Foundation, while its open-source projects generate more than 20 million downloads each month. The company launched React Native EU in Wrocław, which is now React Universe Conf, and organizes the separate Agent Conf in Warsaw. Mike is also an OpenAI Codex Ambassador. He is also an angel investor, and when he is not working, you are most likely to find him on a race track. - Talk description: Talk description to be announced. ### Michał Pierzchała - Company: Callstack - Talk: “Keynote” - Co-speakers: Mike Grabowski - Links: [LinkedIn](https://www.linkedin.com/in/thymikee/) - Bio: Principal Engineer at Callstack. Passionate about building mobile and web experiences, high quality JS tooling and Open Source. Core React Native Community & Jest contributor. Space exploration enthusiast. - Talk description: Talk description to be announced. ### Nader Dabit - Company: Cognition - Theme: Orchestration & Multi-Agent Systems - Talk: “Navigating the Abundance Era of Software” - Links: [X](https://x.com/dabit3) - Bio: I have been building software for over 14 years focusing on everything from AI, cloud, web, blockchain, and cross-platform mobile application development. I also specialize in high impact developer marketing and education. I've spent over 4 years working in the blockchain world, and previously spent over 3 years at AWS where I led developer relations for the Front End Web & Mobile team. - Talk description: Talk description to be announced. ### Tejas Kumar - Company: IBM - Theme: Agentic Engineering Fundamentals - Talk: “The Critical Advantage With AI” - Links: [X](https://x.com/tejask) - Bio: Tejas Kumar is a keynote speaker, best-selling author, and web engineer with over 20 years of experience at companies like Spotify, Vercel, and Xata. He wrote "Fluent React" (O'Reilly), a best-selling deep dive into React internals, and hosts the ConTejas Code podcast where he interviews leading voices in web engineering. - Talk description: According to Anthropic CEO Dario Amodei, all entry level jobs will be erased in the next few years. Eventually, this will grow to erase increasing levels of skill as AI becomes more pervasive. This talk uses live coding, illustration, demos, and science to equip developers with empirically validated protocols for success in the present and near term future. ### Anisha Malde - Company: Amazon - Theme: Orchestration & Multi-Agent Systems - Talk: “Agents FC: Harness Your Agents Into a Winning Team” - Co-speakers: Kourtney Meiss - Links: [X](https://twitter.com/anisha_malde), [LinkedIn](https://www.linkedin.com/in/anishamalde/) - Bio: Anisha is a React Native Developer Advocate at Amazon where she is focused on engaging and enabling app developers. When she's not writing guidebooks on developing for TV, you can find her refactoring recipes and savoring bytes across the world - Talk description: Multi-agent orchestration is powerful, but without a defined harness around it, it's a black box. Agents route dynamically, hand off silently, and hallucinate. Harness engineering is about designing the scaffolding around your agents: the environments they work in, how context gets delivered, how outputs get verified, how plans get captured, how the agents coordinate, and where & when humans can intervene. What if you could organise & coach your agents like a World Cup football team? "Agents FC" is our analogy for harness engineering and orchestration as a football team. Each agent role maps to a tested orchestration pattern: the Captain coordinates routing, the Playmaker retrieves context, the Striker delivers the output, the Defenders run mid-flight verification, and the Goalkeeper catches failures at the output boundary before they escape. And then there's the harness: how they pass the ball between each other, what tools each one carries on the pitch, how the team learns from a missed shot, and how every move gets logged (VAR) so you, as the manager, can change formation mid-task to restructure the orchestration, make substitutions when a strategy isn't working, or call a "half-time team talk" to review progress. We'll put the squad on the pitch and walk through what this looks like with a real task and show that, unlike England, Agents FC will 'bring it home' :) ### Kourtney Meiss - Company: Amazon - Theme: Orchestration & Multi-Agent Systems - Talk: “Agents FC: Harness Your Agents Into a Winning Team” - Co-speakers: Anisha Malde - Links: [LinkedIn](https://www.linkedin.com/in/kourtney-meiss/) - Bio: Kourtney is a Developer Advocate at Amazon where she works on improving developer experience across Alexa, apps, and games. She previously worked in developer relations at Box and TraceLink, helping make APIs and SDKs easier to use. Outside of work, Kourtney can be found reading at coffee shops, playing trivia with friends or hanging out with her dog, Leo. - Talk description: Multi-agent orchestration is powerful, but without a defined harness around it, it's a black box. Agents route dynamically, hand off silently, and hallucinate. Harness engineering is about designing the scaffolding around your agents: the environments they work in, how context gets delivered, how outputs get verified, how plans get captured, how the agents coordinate, and where & when humans can intervene. What if you could organise & coach your agents like a World Cup football team? "Agents FC" is our analogy for harness engineering and orchestration as a football team. Each agent role maps to a tested orchestration pattern: the Captain coordinates routing, the Playmaker retrieves context, the Striker delivers the output, the Defenders run mid-flight verification, and the Goalkeeper catches failures at the output boundary before they escape. And then there's the harness: how they pass the ball between each other, what tools each one carries on the pitch, how the team learns from a missed shot, and how every move gets logged (VAR) so you, as the manager, can change formation mid-task to restructure the orchestration, make substitutions when a strategy isn't working, or call a "half-time team talk" to review progress. We'll put the squad on the pitch and walk through what this looks like with a real task and show that, unlike England, Agents FC will 'bring it home' :) ### Chris Trag - Company: Amazon - Talk: “Panel Discussion: Future of Software Development” - Co-speakers: Margaret Pieczkowski, Bartek Pucek, Grzegorz Wolodko, and Carl Ross - Links: [X](https://x.com/chris_trag), [LinkedIn](https://www.linkedin.com/in/ctraganos/) - Bio: Chris is the Head of Developer Evangelism at Amazon Apps & Games, working with developers worldwide across Fire TV, Alexa, Appstore, and Vega OS. He previously built developer programs at Stripe, Evernote, and Roku, and was the web developer for Harvard.edu. trag.dev - Talk description: Frontier models and agentic development upended engineering workflows faster than ever before. Generating code has become the norm and yet architecting durable systems is more challenging than in prior generations. In this closing panel, we'll skip the hype and doom in order to discuss building at the frontier. What has changed in how we build, and what do the next 6, 18, and 36 months look like? How do product teams hire today, how do we grow careers, local vs cloud AI, and who or what decides what gets shipped? Whether you're a longtime developer, an early adopter, or trying to learn these tools for the first time, this session is for you! ### Margaret Pieczkowski - Company: okthink.ai - Talk: “Panel Discussion: Future of Software Development” - Co-speakers: Chris Trag, Bartek Pucek, Grzegorz Wolodko, and Carl Ross - Links: [LinkedIn](https://www.linkedin.com/in/margaret-pieczkowski/) - Bio: Margaret Pieczkowski is co-founder of okthink, where she is building Firehose, the operating layer for engineering teams that want to go dangerously fast. A former engineering manager who now works alongside agents every day, she is focused on what that shift does to engineering teams: how they coordinate and review work, how they are orchestrated, and what the job becomes when writing code turns into directing it. Previously a Senior Engineering Manager at Hinge Health through its 2025 IPO, with earlier roles across startups, scaleups, and a government lab. Her background spans engineering leadership, backend systems, data pipelines, cloud infrastructure and mobile applications. Based in San Francisco. - Talk description: Frontier models and agentic development upended engineering workflows faster than ever before. Generating code has become the norm and yet architecting durable systems is more challenging than in prior generations. In this closing panel, we'll skip the hype and doom in order to discuss building at the frontier. What has changed in how we build, and what do the next 6, 18, and 36 months look like? How do product teams hire today, how do we grow careers, local vs cloud AI, and who or what decides what gets shipped? Whether you're a longtime developer, an early adopter, or trying to learn these tools for the first time, this session is for you! ### Bartek Pucek - Company: The Thinking Company - Talk: “Panel Discussion: Future of Software Development” - Co-speakers: Chris Trag, Margaret Pieczkowski, Grzegorz Wolodko, and Carl Ross - Links: [LinkedIn](https://www.linkedin.com/in/bartekpucek/) - Bio: Bio to be announced. - Talk description: Frontier models and agentic development upended engineering workflows faster than ever before. Generating code has become the norm and yet architecting durable systems is more challenging than in prior generations. In this closing panel, we'll skip the hype and doom in order to discuss building at the frontier. What has changed in how we build, and what do the next 6, 18, and 36 months look like? How do product teams hire today, how do we grow careers, local vs cloud AI, and who or what decides what gets shipped? Whether you're a longtime developer, an early adopter, or trying to learn these tools for the first time, this session is for you! ### Grzegorz Wolodko - Company: Callstack - Talk: “Panel Discussion: Future of Software Development” - Co-speakers: Chris Trag, Margaret Pieczkowski, Bartek Pucek, and Carl Ross - Links: [LinkedIn](https://www.linkedin.com/in/grzegorz-wolodko-capgemini/) - Bio: Grzegorz has spent 24 years on the delivery side of enterprise IT. From programmer-architect in 2002 to running Capgemini's 2,500-person Northern Central Europe organization for clients in Germany, the Nordics and the UK. Currently CDO at Callstack. - Talk description: Frontier models and agentic development upended engineering workflows faster than ever before. Generating code has become the norm and yet architecting durable systems is more challenging than in prior generations. In this closing panel, we'll skip the hype and doom in order to discuss building at the frontier. What has changed in how we build, and what do the next 6, 18, and 36 months look like? How do product teams hire today, how do we grow careers, local vs cloud AI, and who or what decides what gets shipped? Whether you're a longtime developer, an early adopter, or trying to learn these tools for the first time, this session is for you! ### Levan Kvirkvelia - Company: Rork - Theme: Vibe Coding & Development - Talk: Talk to be announced. - Links: [X](https://x.com/levan) - Bio: Bio to be announced. - Talk description: Talk description to be announced. ### Nick Miller - Company: SpaceXAI - Talk: Talk to be announced. - Links: [X](https://x.com/nickwm) - Bio: Nick Miller is an architect and engineer with over 25 years of experience spanning roles as a Senior Software Engineer, Product Manager, CTO, and Founder. Beyond building and scaling his own SaaS platforms, including e-commerce engines that have processed billions of dollars in transactions, he has led specialized engineering consultancies that served as strategic technology partners to global enterprises like Adobe, Amazon, and Nvidia. Nick currently serves as a Founding Field Engineer at Cursor, where he is forward-deployed with customer teams to integrate large language models into real-world production workflows. By working hands-on with some of the world’s most sophisticated engineering organizations, he focuses on the practical application of AI to fundamentally transform how teams build and ship software. - Talk description: Talk description to be announced. ### Kent C. Dodds - Company: EpicWeb.dev - Theme: Agentic Engineering Fundamentals - Talk: “I Built Your Personal Software Ecosystem” - Links: [X](https://x.com/kentcdodds) - Bio: Bio to be announced. - Talk description: What if you had your own npm registry, your own GitHub, your own cron system, your own email address, and your own database? And what if your favorite AI assistant could build software with all of it on your behalf? That's Kody: a personal assistant platform where every user gets a fully isolated software ecosystem. Your AI writes packages into your own git repos, publishes them to your personal package registry with real versioned imports, schedules your jobs, runs your long-lived services and workflows, receives email at your own address, and stores everything in storage only you can touch. I built this entirely on Cloudflare, because Cloudflare is the only place it could be built. This is a technical deep dive into what happens when you take the newest parts of the developer platform to their limits: sandboxing AI-authored user code with Worker Loaders, per-user git hosting with Artifacts, modeling "one of everything per user" with namespaced Durable Objects, and composing D1, KV, R2, Vectorize, Queues, Workflows, and Email Routing into a true multi-tenant platform. If AI agents are going to build and run software for everyone, this is what the infrastructure underneath looks like. You can start building it today. ### David Khourshid - Company: Stately - Theme: Agentic Engineering Fundamentals - Talk: “Goodbye Slop; Welcome Determinism” - Links: [X](https://x.com/DavidKPiano), [LinkedIn](https://www.linkedin.com/in/davidkpiano/) - Bio: David is the founder of Stately.ai and creator of XState, the most popular open-source state machine & statecharts library. He's a longtime advocate for event-driven modeling and visual diagramming as the foundation for reliable UIs and, increasingly, AI agents. When he's not at a computer keyboard, he's at a piano keyboard. - Talk description: Vibe coding feels productive until you have to maintain it. Behind the agent "thinking..." you ignore and the code you never opened lies a growing pile of wasted tokens, nondeterministic behavior, and compounding errors hiding in plain sight. This talk pushes back on the status quo: elaborate agent architectures, "prompting astrology", overnight automation loops burning through context windows and credit cards. We'll look at why determinism, state machines, and explicit guardrails aren't "old-school"... they're the only sustainable way forward. We'll look at why vibe coding compounds technical debt faster than any era before it, and how modeling behavior explicitly lets you ship AI features you actually understand, maintain, and trust. ### Piotr Karwatka - Company: Open Mercato - Theme: Adoption & Case Studies - Talk: “Agentic Coding at scale at Open Mercato” - Links: [X](https://x.com/piotrkarwatka?s=11), [LinkedIn](https://linkedin.com/in/piotrkarwatka) - Bio: Piotr is in software dev for 26 years already, started as developer and still loves to code and build. In his career he successfully founded and exited Divante and Callstack. In the meantime he built the first version of Alokai (former Vue Storefont) and Open Loyalty - product companies with strong grounding in Open Source. Invested in 30+ startups including Eleven Labs in three very first round. Now building Open Mercato. - Talk description: In the speech I’ll cover how do we engineer, manage, ship a large scale (800k LoC) open source ERP Framework - Open Mercato with 100+ active contributors and weekly releases (about 260+ PRs each) using autonomous coding agents and skills for most if not all SDLC stages from spec to automated integration tests. ### Kitze - Company: Sizzy - Theme: Vibe Coding & Development - Talk: “From Vibe Coding to Vibe Engineering” - Links: [X](https://x.com/thekitze) - Bio: Kitze loves to rant about webdev. He is the founder of Sizzy - the browser for developers. He created React Academy to teach web development. He's the creator of Zero To Shipped - the interactive video course for mastering Fullstack Development. He's documenting his journey on YouTube, streams on Twitch, and has a blog that he maintains once in a blue moon. He made Benji, Twizzle, Sotto, Benji, Glink, showGPT. - Talk description: Web development has always moved in cycles of hype, from frameworks to tooling. With the rise of large language models, we're entering a new era of 'vibe coding,' where developers shape software through collaboration with AI rather than syntax. This talk explores what that means for the future of coding, especially in frontend development, and how it echoes the past while redefining what comes next. ### Sven Großmann - Company: Grafana Labs - Theme: Orchestration & Multi-Agent Systems - Talk: “When "Vibes" Aren't Enough: Observability for AI Agents” - Co-speakers: Ivana Huckova - Links: [LinkedIn](https://www.linkedin.com/in/sven-gro%C3%9Fmann-26387311b/) - Bio: Sven Großmann is a Software Engineer on the AI team at Grafana Labs, where he focuses on building Grafana Assistant and pioneering agentic applications. A dedicated hacker and tinkerer at heart, Sven is a frequent participant in live hacking events and a multi-time hackathon winner. - Talk description: Building AI agents is easy. Knowing whether they work in production is the hard part. When we launched Grafana Assistant to thousands of users, we quickly learned that benchmarks and pre-deployment testing could not capture the unpredictability of real-world interactions. Users behave in unexpected ways, agentic workflows introduce hidden complexity, and small changes can silently impact quality, cost, or security. In this session, we’ll go under the hood of Grafana AI Observability, the platform we built to understand how agents perform, behave, and stay secure in production. Drawing from operating Grafana Assistant at scale, we’ll explore three key dimensions of AI observability: 1. Performance: Understand reliability, token usage, cache efficiency, latency, and the hidden costs of agentic workflows 2. Behavior: Continuously verify that responses and actions are accurate and helpful using online evals and real-time conversation analysis. 3. Security: Proactively detect prompt injections and unsafe inputs before they impact systems or users. Whether you’re building one agent or many, this talk will show how observability helps you confidently improve AI systems in production. ### Ivana Huckova - Company: Grafana Labs - Theme: Orchestration & Multi-Agent Systems - Talk: “When "Vibes" Aren't Enough: Observability for AI Agents” - Co-speakers: Sven Großmann - Links: [LinkedIn](https://www.linkedin.com/in/ivana-huckova) - Bio: Ivana Huckova is a senior software engineer at Grafana Labs and she focuses on improving logging experience within Grafana. In her free time and during hack days, she loves to build fun monitoring solutions for various things, such as her sourdough starter, avocado plants or candles. - Talk description: Building AI agents is easy. Knowing whether they work in production is the hard part. When we launched Grafana Assistant to thousands of users, we quickly learned that benchmarks and pre-deployment testing could not capture the unpredictability of real-world interactions. Users behave in unexpected ways, agentic workflows introduce hidden complexity, and small changes can silently impact quality, cost, or security. In this session, we’ll go under the hood of Grafana AI Observability, the platform we built to understand how agents perform, behave, and stay secure in production. Drawing from operating Grafana Assistant at scale, we’ll explore three key dimensions of AI observability: 1. Performance: Understand reliability, token usage, cache efficiency, latency, and the hidden costs of agentic workflows 2. Behavior: Continuously verify that responses and actions are accurate and helpful using online evals and real-time conversation analysis. 3. Security: Proactively detect prompt injections and unsafe inputs before they impact systems or users. Whether you’re building one agent or many, this talk will show how observability helps you confidently improve AI systems in production. ### Jakub Kuzimski - Company: Allegro - Theme: Agentic Engineering Fundamentals - Talk: “Stop Making LLMs Deterministic. Building a Spec-Driven Software Factory at Allegro.” - Links: [LinkedIn](https://www.linkedin.com/in/jakubkuzimski/) - Bio: Jakub Kuzimski is a Principal Software Engineer at Allegro, one of Europe's largest e-commerce platforms, where he's part of the team driving AI-native development, building spec-driven pipelines, code-review agents, and the evaluation infrastructure that makes agent-generated code trustworthy at scale. After nearly a decade at Allegro as an engineer and engineering manager, he now focuses on how teams change when agents handle implementation. - Talk description: We set out to generate e-commerce software from specifications and immediately hit the wall everyone hits: the same prompt never yields the same code. The breakthrough didn't solve that. It realized it was the wrong target. This talk walks through the architecture we built at Allegro for one of Europe's largest e-commerce platforms: why we moved the guarantee from "reproducible generation" to "reproducible behavior against a contract," how we detect what an agent doesn't know it doesn't know, and the verification hierarchy that catches the error coding agents make most. Silently filling a gap with a plausible guess. I'll show the mechanisms that survived contact with design and the ones still waiting for contact with reality. We're mid-pilot. I’ll share what's working, what's frightening, and the one number that will tell us whether the whole thing earns its keep. ### David Mokos - Company: Expo - Theme: Agentic Engineering Fundamentals - Talk: “Your Biggest User Isn't Human Anymore: Designing DevTools for AI Agents” - Links: [X](https://x.com/davidmokos_) - Bio: David is a YC alum, he built SteerCode - the first mobile AI app builder in AppStore which was acquired by Expo. Now, he's building new agentic products in Expo. - Talk description: At Expo, AI agents are quickly becoming our primary users. Today, more than half of all Expo CLI commands are executed by agents and that share is rising. Building for agents means more than adding an MCP server. It reshapes interfaces, documentation, evaluation, and feedback. ### Zuzanna Skoczylas - Company: Sierra - Theme: Agentic Engineering Fundamentals - Talk: “The AI Agent Iceberg” - Links: [LinkedIn](https://www.linkedin.com/in/zuzanna-skoczylas-b22339151/) - Bio: Zuzanna Skoczylas leads AI Engineering for Financial Services at Sierra, building production AI agents for enterprise clients. Sierra's agents now power customer experience for 40% of the Fortune 500. Zuzanna focuses on the hard problems of deploying conversational AI at scale in regulated, high-stakes environments, translating that work into practical lessons on reliability, trust, and real-world agent design. Her talks bridge the gap between AI hype and what it actually takes to ship agents customers depend on. - Talk description: There's a wide gap between an agent that demos well and one that holds up in production - the flashy demo is just the tip of the iceberg. Having launched customer-facing agents for more than 40% of the Fortune 50, Sierra shares three hard-won lessons from building and shipping production agents, inspired by a tech talk at HumanX: 1. Context engineering — why what you put into (and leave out of) a model's context matters more than model choice, and how Sierra structure's state, memory, and tool outputs to keep agents coherent over long, messy conversations. 2. The agent loop — how Sierra employs a constellation of models, each designed to do the specific agent task it's best suited for within the loop. 3. Benchmarking that actually predicts real-world performance — the story behind τ-bench, the benchmark Sierra built to evaluate how well models handle realistic, multi-turn tool-use tasks, and why it's since been adopted across the industry as a standard for measuring model reliability, not just capability. ### Sujith N S - Company: Ebay - Theme: Adoption & Case Studies - Talk: “Agentic Monitoring in Production: How AI Agents Reduced Alert Investigation Time by 70%” - Co-speakers: Leonel Candido da Silva - Links: [LinkedIn](https://www.linkedin.com/in/sujith-n-s-600641a1/) - Bio: - Designed and developed AI autonomous Agents for monitoring and incident triage - Architected and developed an end-to-end monitoring system using Java and Spring Boot. Mentored a few developers who collaborated into the project. Replaced Celery-based workers with a Quartz scheduler for alert rule evaluation. Implemented configurable lookback periods, automated threshold generation using LLMs, and rule-based decision logic to suppress false positives. Introduced decision rules for holiday periods, budget constraints, model forecasting, and metric comparisons, reducing false positive alerts by over 84% starting November 1, 2025. Currently developing agent-based root cause analysis to enhance alert explainability. - Talk description: Modern monitoring systems excel at detecting anomalies, but they still leave engineers with the most time-consuming task: understanding what happened and determining whether an alert requires action. In our production environment, every alert triggered a manual investigation workflow. Engineers had to open multiple dashboards, compare correlated metrics, analyze dimensional breakdowns, search through operational documentation, and manually determine business impact before escalating findings. Although anomaly detection was automated, incident triage remained largely manual. To address this challenge, we built an Agentic Monitoring platform that transforms alerts into actionable investigations. The system combines multiple specialized AI agents that collaborate to analyze alerts, retrieve operational context, perform metric correlation analysis, identify impacted business dimensions, and generate structured investigation summaries. Rather than replacing engineers, the agents act as investigative assistants that significantly reduce the effort required to validate incidents. In this session, I will share the architecture behind our production agents: 1. Diagnosis Agent: The agent retrieves operational documentation and SOPs and identifies relevant troubleshooting procedures to covert them into structured investigation plans 2. Alert Enrichment Agent: Analyzes correlated metric behavior and detects dimension-level anomalies. It then identifies impacted business segments. Generates investigation summaries and remediation guidance We will discuss how these agents are orchestrated, how monitoring data is transformed into agent-friendly context, and how we addressed common challenges such as hallucination prevention, grounding responses with operational data, latency optimization, and trustworthiness in production environments. The session will include architecture diagrams, implementation patterns, operational lessons learned, and quantitative impact metrics. Since deployment, the platform has reduced alert validation time from more than 20 to 30 minutes to less than 5 minutes per incident while significantly reducing dashboard context switching and manual correlation analysis. The result is faster incident response, improved engineer productivity, and more consistent operational decision making. Attendees will benefit with practical patterns for designing agentic systems that augment operational workflows and can be applied to monitoring, observability, incident management, and site reliability engineering ### Leonel Candido da Silva - Company: eBay Inc. - Theme: Adoption & Case Studies - Talk: “Agentic Monitoring in Production: How AI Agents Reduced Alert Investigation Time by 70%” - Co-speakers: Sujith N S - Links: [LinkedIn](https://www.linkedin.com/in/leonelcandidosilva/) - Bio: Leonel Candido da Silva is an engineering leader exploring how agentic AI can reshape enterprise reliability, observability, and operational intelligence. As a Senior Engineering Manager at eBay, he leads teams building platforms that combine data engineering, machine learning, knowledge graphs, and AI agents to monitor business-critical systems, detect anomalies, reduce false positives, and accelerate root-cause discovery. With more than 19 years of experience across global technology, banking, fintech, and e-commerce organizations, Leonel brings a pragmatic perspective on moving Agent AI from experimentation into production-grade platforms that create measurable business impact. - Talk description: Modern monitoring systems excel at detecting anomalies, but they still leave engineers with the most time-consuming task: understanding what happened and determining whether an alert requires action. In our production environment, every alert triggered a manual investigation workflow. Engineers had to open multiple dashboards, compare correlated metrics, analyze dimensional breakdowns, search through operational documentation, and manually determine business impact before escalating findings. Although anomaly detection was automated, incident triage remained largely manual. To address this challenge, we built an Agentic Monitoring platform that transforms alerts into actionable investigations. The system combines multiple specialized AI agents that collaborate to analyze alerts, retrieve operational context, perform metric correlation analysis, identify impacted business dimensions, and generate structured investigation summaries. Rather than replacing engineers, the agents act as investigative assistants that significantly reduce the effort required to validate incidents. In this session, I will share the architecture behind our production agents: 1. Diagnosis Agent: The agent retrieves operational documentation and SOPs and identifies relevant troubleshooting procedures to covert them into structured investigation plans 2. Alert Enrichment Agent: Analyzes correlated metric behavior and detects dimension-level anomalies. It then identifies impacted business segments. Generates investigation summaries and remediation guidance We will discuss how these agents are orchestrated, how monitoring data is transformed into agent-friendly context, and how we addressed common challenges such as hallucination prevention, grounding responses with operational data, latency optimization, and trustworthiness in production environments. The session will include architecture diagrams, implementation patterns, operational lessons learned, and quantitative impact metrics. Since deployment, the platform has reduced alert validation time from more than 20 to 30 minutes to less than 5 minutes per incident while significantly reducing dashboard context switching and manual correlation analysis. The result is faster incident response, improved engineer productivity, and more consistent operational decision making. Attendees will benefit with practical patterns for designing agentic systems that augment operational workflows and can be applied to monitoring, observability, incident management, and site reliability engineering ### Ruben Casas - Company: Postman - Theme: Agentic Engineering Fundamentals - Talk: “Beyond Components: The Future of AI UI” - Links: [X](https://twitter.com/Infoxicador), [LinkedIn](https://www.linkedin.com/in/ruben-casas-17100383/) - Bio: Ruben Casas is a Google Developer Expert and Staff Engineer at Postman. He builds developer tooling, AI agents, and MCP apps, and writes about frontend architecture at scale and AI. Ruben is a podcast guest and has spoken at major conferences, sharing insights from large-scale migrations and performance work. His talks blend hard-won lessons with forward-looking ideas on modern frontend systems and AI-assisted development. - Talk description: Most interactions with large language models began as plain text. Over time, chat interfaces have evolved to include richer forms of interaction. From first party components returned by tools to fully fledged third party MCP apps embedded directly inside the agent. While these advances dramatically improve user experience, they largely remain stuck in a familiar paradigm: invoking pre-made, static components. At the same time, new code-oriented models are becoming exceptionally good at writing frontend code really fast! This unlocks a major opportunity for true generative UI interfaces that are dynamically created and personalised, without sacrificing predictability, performance, or user trust. In this talk, through concrete examples and demos, we’ll explore the emerging spectrum of generative UI and what the future might look like - Static UI templates exposed through tools and resources - Declarative and configuration driven interfaces - Fully generative interfaces authored on the fly by LLMs ### Nishant Gupta - Company: Meta - Theme: Orchestration & Multi-Agent Systems - Talk: “The Agent Control Plane: Managing Thousands of AI Agents Without Losing Control” - Links: [LinkedIn](https://www.linkedin.com/in/nishant1994/), [LinkedIn](https://www.linkedin.com/in/nishantgupta-ai/) - Bio: Introduction I am a Staff Software Engineer and Researcher at Meta, specializing in large-scale distributed systems and applied AI. I am passionate about building reliable, scalable, and intelligent infrastructure that powers the next generation of agentic workflows. With deep expertise spanning large-scale distributed systems, agentic infrastructure, systems architecture, and operational resilience, I focus on solving the hardest problems at the intersection of systems, AI, and real-world execution where theory meets engineering tradeoffs. Within Meta SuperIntelligence Lab (MSL), I have contributed to building agentic infrastructure - systems where AI agents operate within structured distributed environments, interacting with monitoring, scheduling, and feedback loops. My work in this space focuses on: - Evaluation and auditing of AI-driven decision systems in high-stakes production environments - Reliability, safety, and human oversight in autonomous and semi-autonomous systems - Designing feedback mechanisms to align system behavior with user and operational goals - Measuring real-world impact beyond offline metrics Building Elastic Compute Infrastructure at Meta I also built the next-generation of elastic compute infrastructure to increase overall fleet utilization responsible for managing ~30% of Meta’s capacity (tens of millions of servers) across ~20 geo-distributed datacenter saving billions of dollars in Capex. This also involved partnering with VPs across Ads/Whatsapp/IG/Finance/Infra to set multi-year roadmap and strategy for increasing fleet-wide efficiency. Research At Meta, my recent research includes Dynamic Idle Resource Leasing to Safely Oversubscribe Capacity at Scale, where I designed and deployed a production system that improves datacenter utilization by leasing idle capacity while preserving reliability and strict SLO guarantees. This work required building rigorous evaluation frameworks spanning simulation, controlled experimentation, and real-world safety validation - balancing algorithmic optimization with operational risk. The system has delivered measurable infrastructure-efficiency gains at production scale. I have also authored papers with 90+ citations. What I care about I think deeply about how distributed services communicate, self-coordinate, and act with reliability under ambiguity. My work is rooted in understanding latency, correctness, failure modes, and semantic interoperability - not just performance on benchmarks, but real-­world outcomes that matter in production. I’ve led teams and initiatives that: - Architect complex distributed platforms that serve high-availability workloads at scale - Design agentic systems and frameworks that enable coordinated autonomous behavior across services and models - Build operationally robust infrastructure with strong observability, fault tolerance, and graceful degradation - Translate cutting-edge research into developer-ready systems and patterns Education I graduated from University of California, Los Angeles (UCLA) with a Master's in Computer Science in December 2019. At UCLA, my focus area was on building scalable distributed systems leveraging Machine Learning. Ways to collaborate: • Keynotes, conference talks, and technical workshops • Partnerships with AI platforms, developer tools, and education organizations • Advisory and consulting on AI infrastructure and large-scale systems For speaking, partnerships, or advisory inquiries: nishantgupta@g.ucla.edu - Talk description: Most agent frameworks focus on a single question: "How do I build an agent?" Production systems face a different challenge: "How do I manage hundreds or thousands of them?" As organizations move beyond prototypes, agent ecosystems quickly become difficult to operate. Agents call tools they shouldn't, enter reasoning loops, consume excessive resources, execute conflicting actions, and become nearly impossible to debug. In traditional infrastructure, we solved similar problems using control planes for compute, storage, networking, and orchestration. AI agents need their own control plane. In this talk, I'll introduce the concept of an Agent Control Plane—a new architectural layer responsible for governing execution, permissions, observability, evaluation, coordination, and operational safety across large agent ecosystems. We'll explore: Why agent systems fail outside demos Agent lifecycle management Runtime policy enforcement Permission and trust boundaries Agent evaluation pipelines Multi-agent coordination patterns Observability for reasoning and tool execution Human-in-the-loop escalation mechanisms Operating large-scale agent deployments Through real-world examples, attendees will learn how to move beyond isolated agents and build systems capable of safely supporting thousands of autonomous workflows. Key Takeaways Understand the emerging architecture of agent control planes Learn operational patterns for large-scale agent deployments Design safer and more observable agent systems Avoid common multi-agent failure modes Build production-ready agent platforms ### George Nikolaropoulos - Company: Natech - Theme: Adoption & Case Studies - Talk: “Building a Neobank in Under a Year: How Agentic Engineering Helped Us Ship Snappi” - Links: [X](https://twitter.com/gnikolaropoulos), [LinkedIn](https://www.linkedin.com/in/gnikolaropoulos) - Bio: Accomplished executive and AI leader who inspires innovation and drives success and growth. Has a background in banking, insurance, maritime, and marketing research industries. Always interested in product development, tech strategy, and artificial intelligence. Passionate about creating a culture of trust, collaboration, and creativity with a focus on making customers' lives better. Currently focused on growing and leading engineering teams that create products normal people find impossible, deliver on time, and show respect. - Talk description: Everyone talks about AI agents. Few talk about what happens when you use them to build a bank. In this session, I’ll share the story behind Snappi, Greece’s first ECB-licensed neobank, and how the engineering teams at Natech delivered a fully digital banking experience in less than a year. This is not a story about prompts or prototypes. It’s a story about applying agentic engineering in a highly regulated environment: accelerating software delivery, scaling product development, reducing operational friction, and enabling teams to move at startup speed while building banking-grade systems. I’ll share the practices that worked, the mistakes we made, where AI genuinely multiplied engineering output, and where human judgment remained irreplaceable. If you’re wondering what agentic software development looks like beyond demos and hackathons, this is a production story from the trenches. ### Anjaneya Sastry Kappagantu - Company: AT&T - Theme: Agentic Engineering Fundamentals - Talk: “Learning at Inference Time: TTT, TAO, δ-mem — and When to Spend a Gradient” - Links: [LinkedIn](http://www.linkedin.com/in/anjsastry) - Bio: I am a seasoned Solution Architect at AT&T with over 25 years of experience designing and architecting mission-critical systems. Alongside my professional career, I am a dedicated AI educator and researcher exploring the intersection of advanced computational architectures and classical linguistics. My current work focuses on building high-performance local AI environments and optimizing "frontier-class" models on Apple Silicon using the MLX framework. I actively mentor students on integrating agentic AI workflows and developing specialized LLM applications for complex domain analysis. Note: The views and research presented during this session are my own and do not necessarily reflect the views or positions of AT&T. - Talk description: The dominant assumption of the LLM era is that a model's wiring freezes the instant training ends. To adapt mid-session we cram examples into the prompt — paying a "K/V Cache Tax" that bloats local memory and drags token throughput. A wave of 2025–2026 research is shattering that static paradigm, pulling adaptation into the inference loop itself. This 20-minute talk shows three ways to do it on a local desktop with the MLX graph engine — and they differ in what they change: Test-Time Training (TTT): treat internal projection layers as "fast weights," running in-flight gradient steps to lock onto an out-of-distribution token stream — then discarding the adapter. Test-Time Adaptive Optimization (TAO): use reward signals instead of labels to fine-tune locally from your own unlabeled data, baking best-of-N quality into the weights with no test-time penalty. δ-mem (arXiv, May 2026): leave the backbone frozen and instead update an 8×8 associative-memory state, token-by-token via delta-rule learning, that steers attention — memory without context-window bloat or any weight update. Grounding it in enterprise reality: a live local OpenClaw multi-agent harness troubleshooting a simulated incident. You'll watch an MLX sidecar hit an unmapped log format, adapt on the spot across Apple Silicon's unified memory — no host-to-GPU copies, no prompt bloat, total privacy — and carry that adaptation into the next tool call. On top of these three, I'll show a working prototype of my own: a confidence-gated policy that sits above the mechanisms and decides, per interaction, when a correction is worth a gradient — reinforce in context when the model is right, spend the update only when a human says it's wrong. I'll be explicit about what's a published primitive and what's my design proposal. We close with an honest analysis of TTT's latency cost, the speculative-decoding strategies that could absorb it, and why the cloud's multi-tenant state bottleneck makes local edge inference structurally inevitable for adaptive agents. ### David Budáč - Company: ČSOB a.s. - Theme: Agentic Engineering Fundamentals - Talk: “"Has this happened before?" - building a queryable memory of database logs with agents” - Links: [X](https://x.com/davidbudac), [LinkedIn](https://www.linkedin.com/in/david-budac-98889392) - Bio: Working with Oracle Database for 20 years, as DBA, developer and consultant. - Talk description: Every operations team builds up hard-won knowledge about what their logs actually mean, then loses most of it when people move on or simply forget. We're building an agentic system that captures that knowledge automatically and makes it easier for us the query. I would like to introduce our approach to analyzing database and infra logs with agents. Think of it as Karpathy's Wiki idea - but what if the information was harvested from application's operational logs and metrics instead. I'm a DBA at the largest bank in the Czech Republic, running a large Oracle fleet whose logs and metrics already flow into Elasticsearch under a shared schema. This isn't a real-time monitoring solution. It's an asynchronous agent loop that reviews each time window across the whole fleet, picks out noteworthy events, and writes structured notes, deduplicating and enriching its record of events it has seen before. The result is a growing knowledge base of events, how often they recur, and what they meant operationally. A question that used to cost hours of digging, "has this happened before, and what did it turn out to be?", is now much easier to answer. I'll cover the architecture of the agent loop, how we decide what is interesting, how the system keeps its memory current, and which possible improvments we're seeing in the future. ### Carl Ross - Company: OpenAI - Theme: Orchestration & Multi-Agent Systems - Talk: “No Token Budget: What Breaks When Agentic Software Factories Run on Kubernetes” - Panel: “Panel Discussion: Future of Software Development” with Chris Trag, Margaret Pieczkowski, Bartek Pucek, and Grzegorz Wolodko. - Links: [LinkedIn](https://www.linkedin.com/in/carljross/) - Bio: Member of Technical Staff @ OpenAI - Talk description: A technical case study from OpenAI on using Kubernetes and CNCF-stack patterns to operate agentic software factories across the full SDLC when token budget is entirely removed as a constraint. At OpenAI, we explored what happens when agentic software factories process tens of millions of tokens per minute across the full software development lifecycle: design, planning, implementation, testing, review, release, production monitoring, incident triage, and remediation. At this scale, agentic software development stops looking like a chatbot and starts looking like a distributed cloud-native control plane. Agents do not just write code; they explore designs, implement changes, run tests, review diffs, coordinate releases, inspect production signals, explain failures, and propose remediations. The development model changes from sequential human execution to massively parallel, hermetic, continuously evaluated software delivery. This talk shares a technical case study of running these systems on Kubernetes and CNCF ecosystem components: workflow orchestration, event-driven queues, Kubernetes workers, hermetic sandboxes, MCP-style tool gateways, workload identity, policy gates, eval systems, release controls, trace pipelines, production-monitoring integrations, and human approval flows. Once token budget disappears as a constraint, the bottlenecks move to cluster capacity, queue fairness, sandbox isolation, repository contention, flaky tests, release safety, production feedback loops, agent alignment, multi-agent consensus, and human review. Attendees will learn what breaks after tokens stop being scarce, and how to design cloud-native agent platforms that prioritize development velocity without sacrificing isolation, observability, policy, or production safety. ### Artur Morys-Magiera - Company: Callstack - Theme: Adoption & Case Studies - Talk: “Quantization Impact On LLM Performance: A Case Study of Apex” - Co-speakers: Lech Kalinowski - Links: [X](https://x.com/artus9033), [LinkedIn](https://linkedin.com/in/artur-morys) - Bio: React Native & React developer, Senior React Native Developer at Callstack, a PhD candidate at AGH University of Krakow, MEng in Automation & Robotics. Aficionado of programming, robotics, AI & pointillist paintings. - Talk description: Quantization promises smaller, faster models - but how much capability do you give up? This talk starts with a practical introduction to neural network quantization in the context of LLMs: what it is, how it works, and which trade-offs matter. We then put theory to the test with Apex. Running the original model alongside fp16, int8, and int4 variants, we benchmark all of them on React Native Evals and compare the results head to head. The takeaway: a concrete picture of where quantization yields minimal accuracy losses, and where it starts to hurt. ### Lech Kalinowski - Company: Callstack - Theme: Adoption & Case Studies - Talk: “Quantization Impact On LLM Performance: A Case Study of Apex” - Co-speakers: Artur Morys-Magiera - Links: [X](https://x.com/LeSiOO), [LinkedIn](https://www.linkedin.com/in/lech-kalinowski/) - Bio: Dr. Lech Kalinowski is a PhD-trained physicist, Senior AI Systems Engineer, and hands-on builder working across local inference, agentic systems, mobile AI, and physical AI interfaces. His background spans condensed matter physics, printed electronics, data science, blockchain, local LLM infrastructure, and embedded prototyping. At Callstack, he works on practical AI engineering, including local and mobile AI systems, AI-assisted development workflows, React Native Evals, and agentic tooling. He is the author of technical work on context engineering, multi-agent CLI workflows, quantization, and local model deployment. He previously served as CTO at Panamint, where he worked at the intersection of sustainable electronics, nanomaterials, and applied data science. He is currently building Vault OS, an experimental handheld AI terminal that connects ESP32-based hardware, OLED/e-paper UX, local models, backend orchestration, and OpenClaw automation into a physical control surface for personal agents. - Talk description: Quantization promises smaller, faster models - but how much capability do you give up? This talk starts with a practical introduction to neural network quantization in the context of LLMs: what it is, how it works, and which trade-offs matter. We then put theory to the test with Apex. Running the original model alongside fp16, int8, and int4 variants, we benchmark all of them on React Native Evals and compare the results head to head. The takeaway: a concrete picture of where quantization yields minimal accuracy losses, and where it starts to hurt. ### Dom Sipowicz - Company: Vercel - Theme: Vibe Coding & Development - Talk: “Build an Agent with Eve: The Open Agent Framework” - Links: [X](https://x.com/dom_sipowicz), [LinkedIn](https://www.linkedin.com/in/dominiksipowicz/) - Bio: Dom Sipowicz is a Forward Deployed Engineer at Vercel, where he works with global enterprises to build and deploy AI workflows, production-grade agents, and modern web applications on the Vercel AI Cloud. With over 15 years of experience in web architecture, he's led billion-pound e-commerce replatforms and built high-performance Next.js applications focused on scale and Web Performance. - Talk description: Talk description to be announced. ### Ivan Charapanau - Company: Viktor - Theme: Orchestration & Multi-Agent Systems - Talk: “How will we prompt AGI? A History of Harness Hacks.” - Links: [LinkedIn](https://www.linkedin.com/in/ivan-cherepanov/) - Bio: Ivan Charapanau is a Harness Engineer at viktor.com, building the system that runs AI employees in production. A decade of full-cycle engineering, maintainer of Harbor (3k+ stars). Warsaw-based, writes at av.codes. - Talk description: Most workarounds in agent systems compensate for a specific model limitation. When a new release removes that limitation, the workaround usually stays in the code. Over time this leaves scaffolding that adds complexity without adding value and can cause failures of its own. This talk reviews the workarounds we introduced across several model generations, the releases that made each one unnecessary, the cost of keeping them longer than needed, and workarounds we might need in the future. It ends with a test for deciding which scaffolding to keep or remove after each model upgrade. ### Kevin Lewis - Company: Apify - Role: Conference MC - Links: [LinkedIn](https://www.linkedin.com/in/phazonoverload/) - Bio: Kevin is a developer advocate and educator with over a decade of experience in developer tooling, education, and community growth. He helps engineers learn faster, build better workflows, and get more value from the platforms and tools they use every day. He’s a JavaScript developer (often working with Vue), an avid boardgamer, and a tired dad of two amazing kids. He’s based in Berlin despite not drinking beer or liking techno. - Talk description: Kevin is the conference MC. ## Partners - Platinum Partner: [Amazon](https://developer.amazon.com/). - Gold Partner: [Expo](https://expo.dev). - Silver Partners: [TesterArmy](https://tester.army), [OpenAI](https://openai.com), [Revenue Cat](https://www.revenuecat.com), and [Viktor](https://viktor.com). - Partnership: [meet.js](https://meetjs.pl) and [Open Mercato](https://www.openmercato.com). - Community Partners: [AI Tinkers Poland](https://aitinkerers.org), [AI Coding Summit](https://aicodingsummit.com/), [G2i](https://www.g2i.co), [Brave](https://www.brave.courses/), [React Native Rewind](https://thereactnativerewind.com), [The Thinking Company](https://thinking.inc), [startWarsaw](https://www.startwarsaw.org/), [ITCorner](https://itcorner.org.pl/en/), [MLinPL](https://mlinpl.org), and [webAmigos](https://webamigos.pl). - Media Partners: [Ucz Mnie](https://www.uczmnie.pl) and [justjoin.it](https://justjoin.it/). Past partners featured by Agent Conf include ElevenLabs, Vercel, Amazon Developer, Expensify, Zalando, Chase, HelloFresh, MongoDB, Sentry, and Codemagic. ## Workshop - [Build Your Own Software (re)Factory](https://agent.sh/workshop) takes place September 16, 2026 and is currently sold out. - Giovanni Laquidara and Chris Trag lead the hands-on workshop. - Participants take feature requests from PRD to executable tickets, coordinate parallel coding agents, add automated verification and retry gates, and extend a controllable software factory end to end. ## Optional - [Partner deck](https://agent.sh/Agent%20Conf%20%2726.pdf): Sponsorship information for potential Agent Conf partners.