Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
June 17, 2026
·
Berlin
Lanch: RL Retrieval Agents
Learn how to build and fine-tune custom retrieval agents using GRPO and on-policy distillation for efficient, specific, and faster AI responses.
Overview
We build a custom retrieval agent for Lanch based on GRPO and On-Policy Distillation. We are working on custom fine-tuning and continuous learning from production feedback.
Demo can show how much more efficient a fine-tuned retrieval agent is vs. generic model. Much faster and much more specific. We can show this side by side live.
Links
Construct Labs researches workload-specific, continually learning AI models embedded systems.
Tech stack