I build AI products, agentic systems, and machine learning applications.
I've spent 10+ years working across data science, machine learning, experimentation, and production AI. More recently, I'm focused on how LLMs, agents, tools, and software systems can be combined to solve real problems reliably.
Based in Singapore.
xiaonan.co · Writing — nanx.cc
- AI agents — tool use, planning, orchestration, memory, and evaluation
- LLM systems — retrieval, classification, fine-tuning, evaluation, and production workflows
- AI-native products — turning model capabilities into useful end-to-end applications
- Machine learning systems — NLP, recommendation, experimentation, and production ML
- Developer tooling — exploring how coding agents change the way software gets built
I'm gradually rebuilding my public GitHub around a smaller set of projects that reflect what I'm working on today.
Experiments and applications around tool-using agents, workflows, evaluation, and reliable execution.
Practical implementations involving LLM classification, retrieval, evaluation, fine-tuning, and inference.
Selected work across NLP, computer vision, recommendation, reinforcement learning, and applied ML.
Smaller implementations, prototypes, and experiments documenting things I'm learning or testing.
More projects coming as I move more of my current AI work into public repositories.
My work has covered:
AI & Machine Learning
LLMs · NLP · Recommendation · Ranking · Experimentation · Model Evaluation
Data & ML Systems
Python · SQL · Spark · Airflow · BigQuery · MLOps
Application Engineering
FastAPI · TypeScript · Next.js · PostgreSQL · Docker · Cloud
I'm particularly interested in the boundary between models, software systems, and product decisions — where AI becomes something people actually use.
I write about AI engineering, machine learning, data science, experimentation, and things I learn while building.
Read my technical notes and essays →
For a broader overview of my work:
Building and writing in public around AI agents and AI-native systems.
If you're working on something interesting in this space, feel free to reach out.