2026 Schedule
Tuesday, December 15th, 2026
Badge Pick-up & Light Continental Breakfast
Conference Introduction and Keynote:
When Software Becomes a User: Identity and Authorization for Agents in Production
Break
Supercomputers and AI Agents for Self-Remediation at Azure High Performance Computing
A year ago, operating our GPU fleet scaled O(n): more clusters meant more humans running more runbooks. Today a control plane manages AI-supercomputer GPU capacity across US data centers with near-constant human effort.
One Infrastructure, Every Model: How Netflix Scales ML & GenAI
Building and scaling a machine learning inference platform that serves dozens of use cases- from homepage personalization to fraud detection- is as much an organizational challenge as a technical one.
The Future of the Engineering Org
It’s here. It’s just not evenly distributed.
Break
How AI Reshapes Organizations: Lessons from Enterprise Adoption
What changes when AI moves from an employee tool to a system that carries out work across the organization?
How To Orchestrate, Govern, and Operate AI Coding Agents
Adopting AI coding agents is easy to start: install a tool, open a terminal, and give it a task. Making those agents part of a shared engineering workflow introduces broader challenges. Which agent should handle each task? Where should it execute?
The New Architecture for Generative Search and Recommendations
Search and recommendation systems have traditionally retrieved candidates from an index and scored them independently.
Lunch
Building in the AI-First World: Scale Brilliance, Not Bottlenecks
In an AI-first world, everyone can be a builder. But building is just the beginning. Shipping and running reliable applications that users love is what matters. This session will reveal a critical truth: AI's primary role is that of an amplifier.
High-Signal AI Code Review That Adapts to Your Codebase at Scale
AI coding tools are generating more code, faster, but human code review remains a bottleneck.
Break
Beyond the Sandbox: Provisioning and Securing Production AI Agents at Figma
At Figma, AI agents generate code, call tools, and continue working over long-lived connections. We built Nimbus to provision isolated execution environments for them, but learned that the production trust boundary extends beyond the sandbox.
Make Each Token Count: Compression, Caching, and Retrieval for Production LLM Context
Every agentic AI system today is running the equivalent of an uncompressed MapReduce shuffle: bloated contexts, redundant retrievals, and cache misses that silently multiply cost and latency.
Break
Agent Evals Your Whole Org Can Write and Keep Running
Your vibe eval passed. You ran it again and it failed. Now what?
The Software Factory: How a Swarm of Open Models Can Out-Code the Frontier
Conventional wisdom says a software factory should run on frontier models - which are also the most expensive and typically closed-source. This talk challenges that.
Break
Bring the Power of On-Device AI to Life with Google AI Edge and Gemma
Google AI Edge stacks like LiteRT have powered massive Google and third-party edge applications globally, from leading apps like YouTube and Google Meet, to platforms like Android, Pixel and Chrome.
Building an AgentOps Stack for Financial Services
Agents in financial services carry a constraint most agent stacks never have to meet: sooner or later every one of them has to explain itself — to an owner, an auditor, or a risk committee.
With Great Tooling Comes Great Responsibilities
AI tooling for coding, productivity, and as a product is taking over the engineering world but with all this great tooling, some questions remain. How do you keep track of your spend and ensure that ROI is clear? How do you secure your system and your data? How do you govern your agents?
Attendee Reception - Dry Snacks & Drinks
Wednesday, December 16th, 2026
Badge Pick-up & Light Continental Breakfast
Conference Introduction and Keynote:
Audio Processing with Large Language Models
Break
Making AI Adoption Work Across an Engineering Team
Giving every engineer access to coding agents is only the beginning of adoption. One person may discover a useful way of working while teammates struggle to reproduce it, lack time to experiment, or find that it does not fit their tasks.
Inside LinkedIn's Kubernetes Ops Agent: Skills, Tools, and Guardrails
An AI agent that explains a failed deployment is useful. An agent that can help operate LinkedIn's Kubernetes-based compute platform spanning 500,000+ nodes and 5M+ pods needs more than a good prompt.
Building Pinterest’s VLM Serving Stack on NVIDIA Dynamo
Pinterest has engineered a high-performance Vision-Language Model (VLM) serving stack built on NVIDIA Dynamo, designed to resolve the significant challenges of multimodal workloads, including heavy prefill computation and large KV cache pressure.
Break
From Speech Models to Voice Agents
Voice agents bring natural conversation to everyday workflows by coordinating three distinct workloads: speech recognition processes continuous audio, a language model reasons and calls tools, and speech synthesis generates audio that must keep pace with playback.
Navigating Agentic Standards: What to Build On Now and What to Watch
Teams putting agents into production keep meeting the same pile of names: WebBotAuth, OAuth, WIF, WAG, ID-JAG, MCP, A2A, AAuth, PACT/MoLE, and at least seven payment protocols.
Be Your Toughest Customer: Shipping Collaborative Agents Twice, to Our Own Org and to Lawyers
Harvey has two demanding customers: lawyers, and ourselves. Before our agents touched a legal matter, we shipped them internally as Spectre, our cloud agent platform, now past 7,000 merged pull requests across every repository.
Lunch
Reflections on the 2x Initiative at Honeycomb
Between early 2025 and April 2026, peak weekday PR merges to Honeycomb's main monorepo went from about 30 to about 74, and have held steady in the months since. Most of the delta came in the form of PRs that we could prove AI touched.
Agents That Improve Over Time: Lessons from Shipping an Optimizer
Improving an agent today usually means a manual loop: observe failures, hand-edit the agent's configuration, then re-run evals to validate the fix. This works for one agent, but becomes untenable for a fleet.
Break
Building Agents for Production: Lessons from a LangChain SRE Agent in Production
AI agents are easy to prototype but notoriously difficult to harden for production. This talk deconstructs the architecture of an autonomous SRE agent built on LangGraph and Deep Agents.
Open Source Agentic Inference: Unlocking the Harness with vLLM
Coding agents are turning inference into a continuous workload. A single developer task can trigger dozens of model calls, tool executions, and an expanding context window.
Break
Session-Tenant Compute & Sandbox Density with AX & Substrate
An agent's unit of work is not a stateless HTTP request. It is a session: long-running, stateful across both memory and the local filesystem, and idle 50% to 95% of its lifecycle while waiting on LLM token generation or a human approval click.
From Intent to Enforcement: Open-Source Evaluation and Governance for AI AgentsPresentation
AI agents are increasingly entrusted with complex tasks, sensitive data, and real-world actions...but most teams still evaluate and govern them using disconnected tools, generic benchmarks, and controls tied to a single platform or framework.
Break
From Signal To Impact: The Eval Loop Behind Doordash’s Agentic Ordering System
Running an LLM judge is the easy part of agent evaluation. The harder problem is building a system whose results are repeatable, diagnostic, and actionable. Production agents operate across multi-turn conversations, tool calls, specialized agents, and changing external state.
A Software Factory Build Report: What Worked, What Broke, What I'd Do Again
Software factories aren't a research demo. At Tessl we run one that merges over 1,000 PRs a week, as well as maintenance agents that raise their own issues on the factory itself, and creates changes that merge without us ever knowing they exist.
Building On-Device Health Models with Small, Medium, and Massive Datasets
Most models never make it from the notebook to the device - they're too big, too slow, too power-hungry, or trained on data that doesn't look like what the sensor actually sees.
QCon AI Closing Reception - Dry Snacks & Drinks