Overview
INFER/CON is a one-day conference for the people building and operating the AI stack. Its premise is that as AI moves from experimentation into core production infrastructure, companies still tend to solve parts of the problem in isolation. The conference brings those pieces together to explore how AI systems can run reliably, efficiently, and securely, and what happens when inference operates at global scale.
The event takes place on Monday, November 16, 2026, in San Francisco, California, from 8:00 AM to 6:00 PM PST. It is convened by Datadog. Attendance is free but by application only, and all applications are subject to approval. Submitting an application does not guarantee admission.
Who should attend
The conference is aimed at AI engineers, researchers, founders, and technical leaders working on inference, AI infrastructure, and production AI systems. The page presents it as a chance to step away from day-to-day work, compare approaches with peers, and bring a team's toughest inference questions into the room.
Format and scale
The event expects about 350 people in the room, 20 speakers across the AI stack, and 22 technical sessions. Some sessions will be recorded, but not all. Questions can be sent to team@infercon.ai.
Topics and tracks
The program is organized around several technical themes:
- Production inference, including latency, throughput, scale, and GPU reliability
- KV cache, memory, and disaggregated systems
- Edge, cloud, and heterogeneous compute
- Evaluation, observability, and benchmarking
- Security, safety, privacy, and inference economics
Speakers
The lineup includes co-founders, CTOs, chief scientists, and engineering leaders from inference platforms, AI labs, and cloud providers. Named speakers include Simon Mo of Inferact, co-founder and co-lead of the vLLM community effort; Katelyn Lesse, Head of Platform Engineering at Anthropic; Dmytro "Dima" Dzhulgakov, CTO and co-founder of Fireworks AI; Tri Dao, Chief Scientist at Together AI; Alexis Lê-Quôc, Co-founder and CTO at Datadog; Nikola Borisov, CEO and co-founder of DeepInfra; Banghua Zhu, co-founder and CTO of RadixArk, which builds SGLang; Ryan Hanrui Wang, SVP of Applied AI at Nebius; Nancy Zheng, VP of Engineering of AI Inference at Cerebras; Natalie Serrino and Omid Azizi of Gimlet Labs; Sami Jaghouar, Head of Research at Prime Intellect; Charles Frye of Modal; Ali Taha of Baseten; and Nick Khami of Morph. The organizers note that more speakers will be announced.
Bringing value back to your team
The page encourages attendees to agree on the questions they want to investigate before arriving, compare notes with fellow engineers, and return with a short brief, one prioritized experiment, and clear next steps. It also provides a draft approval note that can be adapted to request time and budget. The suggested plan includes agreeing on three core questions beforehand, capturing three peer benchmarks and two workflows to evaluate during the day, sharing a one-page brief in the first week back, and reviewing the experiment against a baseline after one month.
FAQ
What is INFER/CON? A one-day conference for people building and operating the AI stack, focused on running AI reliably, efficiently, and securely at scale.
Is it free? Yes, but attendance is by application only and subject to approval.
How do I apply? Use the "Apply to Attend" button on the website.
Will sessions be recorded? Some sessions will be recorded, but not all.
