Backend systems × AI-native workflows × products that ship
I build practical software for developers, knowledge workers, and teams that would rather automate the boring parts.
中文 · English · 日本語 · 한국어 · Français
| Area | What that means in practice |
|---|---|
| AI-native developer workflows | Spec-to-ship tooling that turns fuzzy requirements into plans, implementation, self-testing, and review. |
| Enterprise systems & automation | Backend services, low-code platforms, and internal tools that make complex operational work more reliable. |
| Focused, local-first products | Small tools and product experiments that stay fast, private, and useful without unnecessary infrastructure. |
| Direction | Current focus |
|---|---|
| Agent systems | Tool use, orchestration, context engineering, evaluation, and workflows that remain dependable beyond a single prompt. |
| Lightweight & domain-specific models | Following, studying, and applying compact or vertical models — where specialization matters more than raw scale. |
| Retrieval & knowledge systems | Using RAG selectively as one component of a larger system, with more attention to grounding, traceability, and evaluation than to retrieval alone. |
Systems & backend — enterprise SaaS, low-code platforms, distributed services, and the unglamorous work of making them dependable.
AI & model systems — agents, model integration, evaluation, local inference, and retrieval where it adds real value.
