SF Tech Week Private Dinner & Fireside Chat: Beyond the Base Model | Post-Training, Evals & Agentic Systems

Prolific × AI Circle are bringing together an intimate group of just 20 researchers, engineers, and technical leaders for a private dinner and fireside conversation exploring what comes next.

We’ll dig into how teams are thinking about post-training today, where human and automated evaluation each add signal, what we can learn from full agent trajectories rather than final outcomes alone, and what it would actually take to build systems that can learn from their own failures and improve over time.

The conversation will explore questions like; When does an eval become training signal? What kinds of feedback actually help an agent improve? What do human experts still see that automated evaluators miss? And what would a genuinely self-improving agentic system need around the model to work?

Leading the fireside chat will be Croby Roset an Applied Scientist at Microsoft Research, whose work spans large language models, search, evaluation.

This will be a small, candid conversation over dinner; no stage, no pitches, and only 20 seats designed to give everyone at the table space to contribute and connect.

____________________________________

About the hosts:
Prolific powers high-quality human data for model evaluation and training. Learn more about Prolific.

AI Circle curates high-agency operators across research, startups, and enterprise. Learn more about AI Circle.