Efficient LLM Inference for Agents: Workshop by Nebius
November 9th, 17:00-21:00 · Tel Aviv · with Token Factory, Tavily and Inferize
Join Nebius Academy technical workshop for AI builders and engineers.
Work with Nebius, Token Factory and Tavily experts on production-grade LLM inference, with a dedicated exam room where attendees can sit an in-person exam and receive a professional certification. Learn first-hand from the Inferize team about scaling inference without paying for idle GPUs.
What's on the agenda
1. Workshop 18:00-19:30 | LLM inference for agentic workflow using Nebius Token Factory and Tavily
Speakers:
- Daria Balashova, AI Engineer, Nebius
- Yonatan Lourie, AI Researcher, Tavily
Take an AI agent from a working demo to a production-ready application using Nebius Token Factory and Tavily. Explore how to combine reliable LLM inference with web search to build agents that can find and use relevant information.
What you'll learn in this hands-on session:
Model selection: How to choose the right model for your use case
Evaluation: How to test your agent and measure performance
Cost: How to assess inference costs and trade-offs
Safety: How to keep your agent safe and reliable
Scaling: How to run agents reliably at scale
2. Talk 19:45-20:00 | From Minutes to Seconds: Rethinking Cold Starts in LLM Inference
Loading large models takes minutes, so teams keep GPUs warm and pay for capacity they don’t use. Guy Bortnikov, Co-Founder and Director of Engineering at Inferize (now part of Nebius Token Factory), will show how Inferize tackles cold starts to make autoscaling fast, and what that means for the cost and reliability of production workloads.
3. Certification room
In-person certification exams in a dedicated exam room, where attendees can sit an exam and receive a professional certification with special swag and rewards for completing the certification onsite (details to follow).
Certification window #1: 17:00-18:00
Certification window #2: 20:00-21:00
To take the exam, bring your laptop and a valid photo ID.