Join us for our AI Show & Tell monthly event. We'll have several fantastic talks, snacks & drinks, in a beautiful venue. Can't wait to see you there!
This event is part of Global AI Community https://globalai.community/chapters/san-francisco/
Agenda
5:30 PM: Doors open6:00 - 6:15 PM: Welcome5 min: Intro by the host, Cedric Vidal, Principal AI Advocate @ Microsoft15 min: Talk - Brian Douglas, Co-Founder @ Paper Compute15 min: Talk - Patrick Joubert, Founder & CEO @ RippleTide15 min: Talk - Sammy Deprez, Founder @ Ping: The Scheduling Layer for AI Agents15 min: Talk - Eddie Aftandilian, Principal researcher @ GitHub Next25 min: Panel + Q&A, moderated by the host, Cedric Vidal7:45 - 8:30 PM: Networking, light bites and refreshments8:30 PM: Event concludes
🎟 Space is limited and entry is strictly first come, first served—arriving early gives you the best chance of getting in. There will be no entry once event reaches capacity. Thank you for understanding.
AI Show & Tell is a monthly series designed to bring together enthusiasts, professionals, and experts in the field of AI. Each event features engaging presentations, interactive demos, and networking opportunities, providing a place for attendees to learn about the latest advancements in AI technology and connect with like-minded individuals.
Talks
Talk title: The Decision runtime: Enforcing Agent Behavior Before Execution — Patrick Joubert, Co-Founder & CEO @ Rippletide
Abstract: Most agentic frameworks focus on what agents can do, but enterprises need control over what they will do. This talk introduces the decision runtime: a pre-execution enforcement layer that intercepts agent tool calls before they happen, evaluates them against a Context Graph, and either approves, modifies, or blocks them in real time.I'll walk through the architecture of Rippletide's Context Graph MCP Server, how it captures organizational context, encodes decision rules, and enforces them at the tool-call level across any LLM-powered agent. Live demo included on coding use case. Bio: Co-Founder & CEO @ Rippletide. 3x founder with exits to SAP, CircleCI, and Sopra Steria. Patrick has spent a decade building AI infrastructure at enterprise scale from conversational AI to code intelligence to coding genai. At Rippletide, he's building the decision runtime that sits between LLMs and production systems: the layer that makes enterprise AI agents enforceable, not just capable. Winner, OpenAI Codex Hackathon (judged by Greg Brockman).
Title: GitHub Agentic Workflows: Automating recurring work with agents — Eddie Aftandilian
Abstract: A lot of work around software projects is repetitive, but still important: triaging issues, digging into CI failures, updating docs, keeping tests in good shape, and handling procedural tasks that come up again and again. This talk is about a new way to automate that kind of ongoing work using agents.
With GitHub Agentic Workflows, you define recurring tasks in markdown and run them in GitHub Actions. They can work across repositories and use information beyond the repo itself, which makes them useful for more than just code changes.
I’ll show what this looks like in practice, build a workflow live, and demo a few examples of the kinds of tasks it can take on.
Bio: Eddie Aftandilian is a Principal Researcher at GitHub Next, GitHub’s R&D group that explores and prototypes longer-horizon developer experiences. He is one of the creators of GitHub Copilot and previously worked on Copilot Workspace, an early agentic development tool. He now works on Agentic Workflows, building continuous agents that run on GitHub Actions to help teams automate development and maintenance tasks across their repositories.
Talk: AI Solution Architect — Sammy Deprez, Founder @ Ping
Abstract: Every AI product will need recurring workflows, not just one-shot prompts.Ping lets agents schedule jobs, trigger secure webhooks, and run reliably across time zones.We’re building the infrastructure that turns AI agents into always-on products.
Bio: Sammy Deprez is an independent AI Solution Architect and builder with 10+ years in AI and data systems.He specializes in turning powerful but messy AI technology into reliable infrastructure.His work focuses on applied AI, scalable architectures, and developer tools.
Talk: Self Healing Pokémon Master
Abstract: I built an agent to speed-run Pokémon Red, and it initially forgot everything between runs. It turns out self-healing feedback loops are great for JRPGs, and I created a telemetry loop for logs to become patterns and patterns to become long-term memory. Each generation gets smarter, making me one step closer to becoming a vibe-coded Pokémon master. Come watch an agent learn from its own mistakes.