events

Building self-improving AI systems

Amit Sharma, who spent a decade at Microsoft Research working on AI reasoning and causal inference and co-founded the open-source PyWhy ecosystem, gives a talk on self-improving AI systems. He splits self-improvement into three axes — envir

Starts

2026-09-18T13:00:00+00:00

Ends

2026-09-18T14:00:00+00:00

Organizer

Lossfunk Event Calendar

Location

Bengaluru, India

Amit Sharma, who spent a decade at Microsoft Research working on AI reasoning and causal inference and co-founded the open-source PyWhy ecosystem, gives a talk on self-improving AI systems. He splits self-improvement into three axes — environment generation, RL training, and harness evolution — and presents two frameworks: executable counterfactuals for generating verifiable RL environments with controllable difficulty, and interwhen, which compiles natural-language policies into runtime verifiers for an agent's reasoning steps. The talk ends on an open question about training models and harnesses jointly. Attendance is in person in Bengaluru or online.