Field Notes

// Transmissions

Working notes from the lab — reinforcement learning, agent evaluation, retrieval systems, and the unglamorous infrastructure that decides whether any of it survives contact with production.

3 entries · written by Karan Bista

003 // Aug 6, 2026· 18 min readNew
Implicit Q-Learning: Offline RL Without Trusting Your Policy
IQL keeps out-of-distribution actions out of the Bellman backup entirely — an expectile value function, a policy-free critic target, and advantage-weighted cloning. Why that restraint is the whole trick.
Reinforcement LearningOffline RLImplicit Q-LearningActor-CriticValue ApproximationAdvantage-Weighted Regression
—views—likes—commentsRead note →
002 // Aug 10, 2024· 4 min read
Building Reliable RAG Systems
Design patterns, evaluation strategies, and latency tips for production RAG.
RAGLLMsVector DB
—views—likes—commentsRead note →
001 // May 2, 2024· 4 min read
From Prototypes to MLOps
A roadmap for turning notebooks into maintainable production systems.
MLOpsCI/CDMonitoring
—views—likes—commentsRead note →

Implicit Q-Learning: Offline RL Without Trusting Your Policy

Building Reliable RAG Systems

From Prototypes to MLOps