ML Engineer building production trading systems and LLM applications.
Two Korean-language books on verifying AI-assisted software, with substantial free editions on WikiDocs:
The full paid PDF editions are in preparation. Core concepts are explained in the text with worked examples and verification criteria; access to private production repositories is not required to understand them.
한국어 책 소개와 읽는 순서 · English introduction
Stock Trading AI — End-to-end ML pipeline processing 26M daily features across 580 stocks
- Reinforcement Learning agents (SAC, PPO, QR-DQN) for pair trading entry/exit
- Ensemble system: 19 GICS sector groups, HMM regime detection, Thompson Sampling voting
- Backtest: Sharpe 1.97, MDD 4.93% (275-day OOS)
- MLflow registry, torch.jit.script 2.13x speedup, FastAPI hot reload serving
ChatBout AI — LangGraph RAG + Multi-Agent System
- 8-node RAG pipeline: HyDE query expansion, Cross-Encoder reranking, Self-RAG hallucination guard
- Supervisor + 4 specialist agents with multi-hop chaining
- RL Thompson Sampling retriever bandit (FAISS / BM25 / Hybrid auto-selection)
- RAGAS evaluation, LangSmith tracing, Docker deployment
ReadBooks.ai — PDF Translation Desktop App
- Tauri 2 (Rust) + SvelteKit, Claude API, 30+ languages, SSE streaming
- Models: huggingface.co/tedpark — QR-DQN 51-quantile pair trading agent
- Blog: tedpark.github.io/blog — 8 technical posts (EN/KO/JA) on RL, MLOps, RAG
- Portfolio: tedpark.github.io
Python PyTorch LangGraph LangChain FAISS FastAPI MLflow Docker
TypeScript SvelteKit Next.js Rust Tauri DuckDB MongoDB Redis