I am Xinhe Liu (Charles), an engineer and technical writer working across data, AI, and product systems.
Essays, technical notes, and evolving books from my work and learning in data, AI, and coding agents.
12
posts
5
notes
4
clusters
Writing activity
6 writing days · 6 pieces created · 12 updates
2026-04-03 → 2026-10-03
Apr
May
Jun
Jul
Aug
Sep
Oct
LessMore
Essays, notes, and book chapters, dated by their original creation and revision.
Writing in this period
- UpdatedContext Management
- UpdatedHow Coding Agents Work
Writing & Books
Complete books, intros, and recent essays share one entry point from the home page.
Computational Advertising (3) Ad Strategy
Strategy notes on budget pacing, traffic allocation, behavior decomposition, and optimization controls inside ads systems.
Recommender Systems 101 (3) Algorithmic Frontiers
A survey of advanced recommendation models, multi-objective optimization, representation learning, and real-time system evolution.
A/B Testing (2) Challenges in A/B Testing
A deeper look at difficult A/B testing cases such as interference, long-term effects, triggering, surrogate metrics, and cluster design.
Causal Inference (1) What is Causal Inference
An introduction to causal inference, including confounding, selection bias, counterfactual thinking, and the role of randomized experiments.
Project threads
Public project notes organized around problem types, system boundaries, and technical judgment.
Growth and measurement systems
Translating ambiguous growth goals into metrics, experiments, launch criteria, and operating feedback.
Recommendation and ads infrastructure
Ranking, retrieval, allocation, and measurement layers that balance user value, marketplace constraints, and business goals.
Causal measurement and experimentation
Separating lift from correlation with holdouts, attribution windows, and counterfactual reasoning.
Enterprise AI and agent workflows
Putting evaluation, escalation, human review, and governance contracts ahead of workflow automation.
About and contact
Professional background, publication boundaries, GitHub, and email in one place.