Books
Technical Publications
Some of the topics I study and work with also become books. Here you'll find my publications available on Amazon.
RLHF and Language Model Evaluation: From Mathematical Theory to Production Monitoring
Master the complete lifecycle of modern language model alignment. This book provides a practical and rigorous guide to RLHF (Reinforcement Learning from Human Feedback), DPO, Constitutional AI, RLAIF, and modern LLM evaluation. Covering both theory and implementation, you'll learn how alignment systems are built, optimized, evaluated, and monitored in production environments.
- Mathematical foundations of RLHF, PPO, reward modeling, and preference optimization
- Modern alignment methods including DPO, KTO, ORPO, Constitutional AI, and RLAIF
- Hands-on implementations using the Hugging Face ecosystem
- Multidimensional evaluation of toxicity, bias, factuality, helpfulness, and harmlessness
- Production monitoring, guardrails, observability, and drift detection