From training to production: Building and scaling real-world LLMs
Building and scaling LLM systems means making difficult engineering decisions and tradeoffs across training, adaptation, inference and deployment.
Join us in London for a technical evening tracing the LLM lifecycle from pre-training and post-training to inference and production use.
Speakers from Nebius, Red Hat, Poolside, SpinnerAI and Cohere will share practical insights and lessons from reinforcement learning, large-scale training, enterprise post-training, and inference optimization for production systems and self-improving agents.
Five technical talks, followed by discussion and networking with speakers and attendees over beers and bites.
Space is limited. RSVP to secure your spot!
Agenda
18:00 — Doors open & networking
18:15 — Welcome from Nebius
18:20 — Reinforcement Learning for LLMs in Practice
Alexander Golubev, Research Lead, Nebius
18:45 — Lessons with large-scale training, scaling laws, context extension
George Grigorev, LLM Pre-training Expert, Poolside
19:10 — Break, drinks & bites
19:40 — Post-training LLMs for enterprise use: bridging the gap between general-purpose models and real-world deployment
Xiaolu Lu, Machine Learning Engineer, Cohere
20:05 — Inference optimization
Eldar Kurtic, Inference Research Lead, vLLM, Red Hat
20:30 — How to decrease your inference costs ×-times via a Unified Knowledge Layer for self-improving agents
Kir Zharov, Founder, CEO, SpinnerAI