Chip Huyen
Contents
Books in this wiki: Ai Engineering, Designing Machine Learning Systems
Background
Chip Huyen is a Vietnamese-American author, lecturer, and ML systems engineer. She has worked at NVIDIA, Snorkel AI, and co-founded Claypot AI (acquired). She is an adjunct lecturer at Stanford University, where she created and teaches the course CS 329S: Machine Learning Systems Design. She has authored several widely-read online resources on ML engineering and MLOps, and her 2022 book Designing Machine Learning Systems (O'Reilly) became a standard reference for ML practitioners building production systems.
Core Positions
Huyen consistently argues that the engineering and operational aspects of ML — evaluation, data quality, monitoring, serving infrastructure — are systematically undervalued relative to modelling and algorithmic work. She is sceptical of hype and benchmarks that do not reflect real production conditions, and places evaluation at the centre of any rigorous AI development process.
In AI Engineering, she extends this perspective to the foundation model era: the shift from training models to adapting them does not reduce the importance of rigour — it changes where that rigour must be applied (evaluation, prompt design, dataset curation, inference cost).
Books
Designing Machine Learning Systems — Designing Machine Learning Systems (2022)
A holistic guide to building and operating traditional ML systems in production. Covers the full lifecycle: data engineering, training data curation, feature engineering, model development, deployment, monitoring, continual learning, and infrastructure. The central argument is that the ML algorithm is a small part of the system — business requirements, data pipelines, deployment, monitoring, and feedback loops are equally important and systematically underemphasised in ML education.
Ai Engineering — AI Engineering (2024)
A comprehensive guide to building production applications on top of foundation models. Covers the full lifecycle: understanding models, evaluating them, adapting via prompt engineering and RAG, finetuning, dataset engineering, inference optimisation, and feedback loop design. Positions evaluation as the most underrated activity in the discipline and treats the probabilistic nature of LLMs as the central challenge all techniques must address. One of the most complete and rigorous books written on the subject.