About
I build applied AI products and write down what the work teaches me.
I am James Li, a computer science student focused on turning technical learning into visible products, case studies, and essays. This site is my public notebook: part portfolio, part lab record, part writing habit.
Work style
Proof before polish.
Applied AI products
I prefer projects with real constraints, inspectable code or case notes, and a clear explanation of what was learned.
Full-stack systems
I prefer projects with real constraints, inspectable code or case notes, and a clear explanation of what was learned.
Machine learning practice
I prefer projects with real constraints, inspectable code or case notes, and a clear explanation of what was learned.
Computer science foundations
I prefer projects with real constraints, inspectable code or case notes, and a clear explanation of what was learned.
Selected projects
A short evidence trail.
AI product / Public case notes, private source
WhyThisMove / IronWall Chess Engine
Built an AI chess coach that combines Stockfish, Maia-style human move analysis, opening training, blunder practice, game review, and LLM explanations.
Full-stack system / Public source available
Drew Rangers Tracker Pro
Built a Vue and Cloudflare system with role-oriented flows for practice, game, admin, health, roster sync, and PDF reporting.
Research AI SaaS / Public case notes, private source
Citely
Built a citation intelligence product using academic APIs, PDF parsing, RAG, citation matching, Supabase auth/storage/RLS, and Stripe credits.
AI education platform / Private case summary
Zhimian AI Interview Platform
Built an AI interview training platform with WebSocket interview sessions, emotion analysis, RAG chatbot support, RBAC administration, and payment proof review.
Technical base
AI Products
LLM coaching, RAG, embeddings, citation audit, repo analysis, interview practice, knowledge graph QA.
Full-Stack Systems
React, Vue, TypeScript, FastAPI, REST APIs, WebSocket flows, auth, RBAC, storage, billing, admin operations.
Data And Infrastructure
PostgreSQL, Supabase, pgvector, ClickHouse, Cloudflare D1, Docker Compose, Caddy, Netlify, Render, Cloudflare Pages.
Testing And Delivery
Vitest, Testing Library, Playwright, pytest, Node test runner, deployment configs, private testing, case-study documentation.
Research and learning
Paper Replication
Reproduced ML workflows such as battery SOH prediction with EIS features, XGBoost, LOOCV, and SHAP analysis.
Kaggle Practice
Built reproducible data science notebooks for feature engineering, model training, validation, and submissions.
Algorithms And CS Foundations
Maintained multi-language algorithm practice, Linux notes, cloud concepts, architecture, databases, networks, and operating systems.