AI Engineer · Agent & LLM · Creator of Skills Manager (3.5K+ ⭐)
I build AI agents and LLM products for real users — on top of production CV/ML systems I've shipped before.
AI Engineer based in Canberra, Australia. Ex-Tencent YouTu Senior Researcher, PhD in Machine Learning, ICCV LVIS Challenge champion. I ship agent tooling and LLM products end-to-end: open-source desktop apps used by thousands of developers, and consumer AI products with paying / active learners.
🌐 tianliangzhang.com · 🎓 Google Scholar
- Agent tooling — central skill libraries, per-agent adapters, presets, and multi-device sync across 50+ coding agents (Claude Code, Codex, Cursor, Copilot, Gemini CLI…)
- LLM products — rubric scoring, structured JSON contracts, eval harnesses, speech pipelines, and 24/7 practice chat — shipped to real users
- Reliable LLM systems — prompt-hash caching, pre-generation, multi-provider fallback, cost/latency budgets under production traffic
- Production CV/ML — sub-second ONNX inference, edge deployment, high-recall industrial inspection (prior work at Tencent YouTu)
- Skills Manager — open-source desktop app that unifies AI-agent skills across 50+ coding tools (Claude Code, Codex, Cursor, Copilot…). One central library, presets, multi-device sync + CLI. Rust/Tauri + React. 3.5K+ ⭐ · 300+ forks
- EasyIELTS — AI IELTS speaking & writing coach: full mock interviewer (TTS → record → STT), 4-dimension rubric scoring, sentence-level feedback. Next.js + Supabase + LLM.
- Sticker It — generative-AI SaaS, photo → custom stickers. Next.js + OpenAI + Stripe + S3
- ACT Learner — bilingual road-rules LLM coach with cached generation + OpenAI/Gemini fallback
- Publications — CVPR 2020 (co-first) · CVPR 2017 · IEEE TITS ×2 (IF 9.55) · ICME 2022
- Competition — ICCV 2021 LVIS Challenge — Rank 1 + Best Innovation Award
- Patents — 2 US · 13 China
- Roles — Senior Researcher @ Tencent YouTu Lab
Engineering deep-dives on what I ship → tianliangzhang.com/blog
Recent topics: central skill libraries for coding agents, making IELTS band scores reproducible with LLMs, eval harnesses that catch silent grader drift, production LLM reliability patterns.
Open to AI Engineer / Agent Engineer roles in Australia.