AI Engineering index
LLM systems, AI system architecture, MLOps, evaluation, deployment, and production AI infrastructure.
Articles on applied AI, data science, product systems, and engineering
A crawlable index of practical notes by Yangming Li, covering LLM systems, AI system architecture, NLP, MLOps, statistical learning, experimentation, and product thinking. The dedicated AI Engineering column collects the posts closest to production AI systems.
Production AI
AI Engineering
RAG evaluation guide with metrics and test sets
How to evaluate retrieval quality, generation faithfulness, citations, no-answer behavior, and production reliability.
Copilot agent golden test set
Build reusable Copilot agent cases, rubrics, schema checks, regression gates, and release decisions.
AI agent evaluation is more important than prompting
Why production AI agents need custom eval sets, trajectory checks, calibrated judges, regression tests, and business-ready metrics.
Testing and evaluating Copilot agents
Schema-first evaluation, custom graders, release gates, human review, and monitoring for reliable Copilot Studio agents.
Model Context Protocol guide
How connected AI tools use external data, tools, and services.
Agentic AI systems with n8n
A technical guide to agent architecture, system integration, and production constraints.
Models and text
Applied AI and ML
A 90% accurate churn model can still lose money
Why high-accuracy churn prediction can improve renewal rate while losing revenue, and how uplift thinking fixes the target.
Beyond A/B Testing: Uplift modeling in industry
A practical guide to CATE, meta-learners, Qini, AUUC, decile lift, online experiments, and incremental ROI.
Causal inference for product analytics
How product teams connect experiments, observational data, ATE, CATE, uplift, guardrails, and decisions.
Model calibration in Python
Reliability diagrams, expected calibration error, confidence bins, decision thresholds, and monitoring checks.
RAG evaluation Python example and scorecard
A layered guide with retrieval metrics, a synthetic golden dataset, framework comparison, and monitoring checks.
AI agent evaluation is more important than prompting
A practical essay on moving agent development from demo-driven iteration to evaluation-driven production readiness.
Testing and evaluating Copilot agents
Moving a Copilot agent from a demo to a reviewed, monitored, production-ready AI system.
Model Context Protocol guide
How connected AI tools use external data, tools, and services.
Agentic AI systems with n8n
A technical guide to agent architecture, system integration, and production constraints.
CMU Advanced NLP course notes
Study notes on foundational and advanced natural language processing topics.
Random Forest explained
An ensemble learning guide with practical machine learning context.
PyTorch review
A compact review of tensors, shapes, broadcasting, reshaping, and distributions.
Trustworthy machine learning
Notes on transparency, fairness, privacy, robustness, and accountability.
Data systems
Data platforms and MLOps
MLOps essential skills
Model management and deployment foundations for production ML work.
Databricks lakehouse guide
Delta Lake, MLflow, Unity Catalog, and data platform implementation notes.
Docker for machine learning
Using Docker for reproducible data science and ML environments.
Kubernetes guide
Container orchestration concepts and implementation patterns.
Product systems
Product and experimentation
A 90% accurate churn model can still lose money
A subscription renewal example showing why coupon campaigns need incremental ROI, holdouts, and uplift-based targeting.
Beyond A/B Testing: Uplift modeling in industry
How to move from average A/B test lift to targeted incremental impact for coupons, campaigns, churn, and growth.
A/B testing sample size in Python
Power analysis, minimum detectable effect, guardrails, and sample size planning for product experiments.
Causal inference for product analytics
When to use experiments, observational causal methods, uplift, and guardrails for product decisions.
High-impact value propositions
A practical framework for defining and validating product value.
What makes products successful?
Notes on user needs, stickiness, and product experience.
A/B test engineering guide
Feature gates, experiments, and reliable launch measurement.
Jira for agile project management
How issue tracking and delivery systems support technical execution.