Han, Not Solo
Essays by Han Lee on machine learning systems, AI evaluation, agent infrastructure, search, recommendation systems, and engineering practice.
AI / MLSoftware / engineering practiceEngineering management / careers
- Posts
- 78 from 26 Jun 2019 to 20 Sep 2026
- Posting
- monthly (12 in the last year)
- Kind
- personal blog
- Feed
- https://leehanchung.github.io/feed.xml
2026 11
- Jev and the return of AI/ML Engineering
- Stop Being Permanent Underclass. Implement a Token Factory.
- Hidden Technical Debt of AI Systems: Agent Evaluation Infrastructure
- Data Aggregation Is Not a Moat
- Hidden Technical Debt of AI Systems: Agent Harness
- Don't Outsource Your Understanding
- Hidden Technical Debt of AI Systems: Agent Runtime
- "Determinism" is the Biggest Cope in AI Adoption
- The AI Great Leap Forward
- A Taxonomy of RL Environments for LLM Agents
- It's-a Me, Agentic AI
2025 20
- Databricks' Strategic Playbook: Reynold Xin on Growth, AI, and the Future of Data Infrastructure
- Claude Agent Skills: A First Principles Deep Dive
- Building for IT 3.0: Why Agent-Native Products Will Crush SaaS Click-Ops
- Enterprise AI Transformation: The 4-Set Framework for IT 3.0
- Statistics for AI/ML, Part 4: pass@k and Unbiased Estimator
- How AI Tools Are Reshaping Software Development Team Responsibilities
- Software Engineering for Data Scientists, Part 1: Pydantic Is All You Need for Poor Performance Spaghetti Code
- No Code, Low Code, Real Code
- MCP is not REST API
- Prompt Deployment Goes Wrong: xAI Grok's obsession with White Genocide
- Agents Are Workflows
- Vibe Coding 101 for Software Engineers
- When Prompt Deployment Goes Wrong: MLOps Lessons from ChatGPT’s 'Sycophantic' Rollback
- Pro Tips: Model Context Protocol developement on Windows
- Walkthrough: Setting Up Claude Code on Amazon Bedrock
- Poking Around Claude Code
- Solve AI Engineering Slop with Machine Learning
- Statistics for AI/ML, Part 3: Measuring Classification Agreements Between Agents
- Statistics for AI/ML, Part 2: Measuring Ranked Agreements Between Agents
- The Differences between Deep Research, Deep Research, and Deep Research
2024 19
- Statistics for AI/ML, Part 1: Adding Confidence Interval to Your Aggregation Statistics
- Reasoning Series, Part 4: Reasoning with Compound AI Systems and Post-Training
- Reasoning Series, Part 3: Reasoning with Prompt Engineering and Sampling
- What The Hype and Reality of Agents: Resurgence of the Past, with LLM Flavors
- Reasoning Series, Part 2: Reasoning and Inference Time Scaling
- Reasoning Series, Part 1: Understanding OpenAI o1
- Updating Your Priors as Compute Scales
- Frameworks for LLMs and Compound AI Systems through the Lens of 50 Years of Semiconductor History
- AI Winter Is Coming
- Seven Key Metrics Every SaaS Business Needs to Thrive in the AI Era
- LLM-as-a-Judge: Rethinking Model-Based Evaluations in Text Generation
- AI System UX: From Chatbots to Agents
- Stop Following Tech Influencers
- First Principles in Evaluating LLM Systems
- NSFW: GPT-4o Tokenizer
- Tools for Large Language Model Agents
- Information Trifecta
- Service as Software
- Subject: search lingo
2023 4
2022 2
2021 14
- Notes on from EMNLP 2021
- AWS Lambda: Keeping Functions Warm
- AWS Sagemaker: Persist Custom Conda Environment
- List All Columns in a Table in Redshift
- Pandas fillna for Categorical Columns
- Syncing Conda Environments with requirements.txt
- XGBoost Hyperparameters Tuning
- SageMaker: Automatically Stop Notebook and Studio Instances When Idle
- Getting Months from a Date Range in Pandas
- Tensorflow and Friends
- Boruta Algorithm
- Weight of Evidence and Information Value
- Managing Multiple AWS and Git profiles
- Feature Selection and Dimensionality Reduction