Skip to content
View tingwei161803's full-sized avatar

Block or report tingwei161803

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
tingwei161803/README.md

Peter (Ting-Wei) Chang, MD | 張庭維

Physician → AI engineer | 醫師轉 AI 工程師

Research Assistant · Institute of Information Science, Academia Sinica · Taipei

Website Publications LinkedIn Scholar ORCID

Two years of residency, three years of shipping LLM systems.
兩年住院醫師,三年做 LLM 系統。


👋 About me / 關於我

Physician-turned-AI engineer with an M.D. and an M.S. in Computer Science (NLP Lab, National Taiwan University). Built an end-to-end AI platform (chatbot + ERP + CRM) for medical clinics at a Taipei AI startup, covering dataflow, inference, serving and monitoring. First author of publications at COLM 2026 and BMJ Health & Care Informatics.

My research sits at the intersection of large language models and clinical medicine: continual LLM adaptation (COLM 2026), medical imaging AI (BMJ Health & Care Informatics 2025), and plain-language adaptation of medical texts (TREC PLABA 2024).

Outside research I read, then build some sites under peteraim.com covering conferences, medical AI, industry reports, startups and security (See Projects). I also co-founded Beyond Medicine (醫外之境), a community empowering healthcare professionals to explore career paths beyond clinical practice.


🔬 Research / 研究

Large language models × clinical medicine. Three first-author papers, three different communities:

COLM 2026 Smarter by the Moment: Environment-Driven Dynamic Policies for Continual LLM Improvement
Ting-Wei Chang, Po-Chun Chen, Hen-Hsen Huang, Hsin-Hsi Chen
Lets an LLM turn its own history into strategies: continual gains, no retraining.
arXiv:2609.16800
BMJ HCI 2025 Artificial intelligence for predicting interstitial fibrosis and tubular atrophy using diagnostic ultrasound imaging and biomarkers
Ting-Wei Chang, Chang-Yu Tsai, Zhen-Yi Tang, et al.
Predicts kidney fibrosis from ultrasound and blood tests, no biopsy needed.
doi:10.1136/bmjhci-2024-101192
TREC 2024
PLABA track
Enhancing Accessibility of Medical Texts through Large Language Model-Driven Plain Language Adaptation
Ting-Wei Chang, Hen-Hsen Huang, Hsin-Hsi Chen
Rewrites dense medical text into language patients can actually read.
arXiv:2609.17398 · NIST SP 1329

All 7 papers with full citations and BibTeX → peteraim.com/publications


⚙️ Engineering / 工程

Founding AI Engineer at Genibuilder(沛智科技), Sep 2024 – Jul 2026. Owned the end-to-end AI platform (chatbot + ERP + CRM) as the company scaled, from first feature to large-scale production:

  • 🏗 End-to-end AI system from scratch: dataflow → inference → serving → monitoring, multi-tenant, designed for scalability
  • 💬 LINE community, follower base up 400%: architected the conversational AI platform behind a single client's community, sustaining thousands of requests per second
  • 📣 AI marketing platform in two months: delivered from scratch, putting AI chatbots into LINE and Facebook Messenger
  • −90% agent response latency: a systematic rewrite of the Genii-S agent architecture, not a caching shortcut
  • 💳 Billing live in one week: subscription and payments end-to-end, the company's first recurring revenue
  • 🔐 ISO/IEC 27001:2022: technical owner defining AI and systems security controls; audit passed Jul 2026

🏅 Awards / 獲獎

Year
2024 AOCR2024 AI Challenge: classification of acute appendicitis on contrast-enhanced CT. Ranked 8/122 in the first stage and 11/25 in the second
2022 Letter of appreciation from the TMUH superintendent: produced 3D-printed organ models for donor body reconstruction after organ retrieval, across 10+ cases at TMUH, Wan Fang and Shuang Ho hospitals
2021 AI CUP 2021 Medical Dialog Analysis Competition: ranked 17/368

🎤 Talks & Workshops / 演講與工作坊

Year
2026 Online talk at 百業分享堂: "Turning long reports into interactive AI-built web pages: a medical-site case study". A live demo of turning the COMPUTEX 2026 site and the Stanford AI Index Report 2026 into shareable interactive pages with AI and no code. Watch the talk
2026 Co-organizer and teaching assistant, Vibe Coding Basics workshop: a volunteer team from the Evolving Intelligence (度化智能) learning community running hands-on AI workshops for non-technical people, open to everyone. First in-person session 22 Aug 2026 in Taipei, 4.58 / 5 overall satisfaction
2025 Online talk at 百業分享堂: "From physician to AI: a career-transition journey". Watch the talk
2022 Invited instructor, medical humanities workshop (3D printing × organ donation), hosted by TMU and Shuang Ho Hospital. Watch the Da Ai TV report

📚 Notes as websites / 讀完就做成網站

I read, then build. 49 bilingual sites under *.peteraim.com, across 10 themes: conferences, medical AI, industry reports, startups, security and more.

🎤 Agentic AI Summit '26 Notes All 147 talks from Berkeley RDI's summit (Aug 1–2, four stages, 14 livestreams): a one-liner, a TL;DR and timestamped links into the livestream for each
🧩 TRIZ 40 Inventive Principles The 40 principles as a bilingual codex, from ancient crafts to AI: 437 cases, 174 sources, flashcards, quizzes
🖥 Computex + Innovex 2026 Highlights from COMPUTEX 2026 and the InnoVEX startup expo, focused on AI integration trends
🩺 AI & Medical AI Daily An automatically curated daily digest of AI and medical-AI news, industry moves, regulation and product analysis
🏥 Taiwan HealthTech Industry Guide Taiwan's health-tech industry in 15 bilingual pages: medical software, digital health & AI, devices, regulation, ecosystem, careers, plus a learning hub for non-medical newcomers
🛡 OWASP Top 10 for LLM Apps 2025 A bilingual reader of the official OWASP Top 10 for LLM Applications 2025
🛠 lazy-html · Data into Web Pages A Claude tool that turns text, lists, files or a topic into a polished, interactive web page, no coding required (repo)
🚀 CS183: Startup, Class Notes Bilingual study notes for Stanford's CS183 "Startup" lectures

Browse all 49 →Conference timeline →


🧰 Stack

Languages Python (uv) · TypeScript · SQL
LLM / Agent Google ADK · LLM Gateway (routing, fallback, rate limiting, cost tracking) · MCP tool registry · RAG · knowledge graphs · multi-agent systems · continual learning
Cloud GCP (Cloud Run, Cloud SQL, Firestore, Redis, Pub/Sub, Vertex AI) · Cloudflare Workers
Data PostgreSQL (row-level security for multi-tenancy) · Firestore · Redis · SQLite
Web Astro · React · Tailwind
Practice GitLab CI/CD · GitHub Actions · canary / blue-green · pre-commit hooks · ISO 27001 controls

✉️ peterchang.ai.med@gmail.com | 📍 Taipei, Taiwan | 🌐 peteraim.com

Pinned Loading

  1. ai-index-report ai-index-report Public

    Five editions (2022–2026) of Stanford HAI's AI Index Report, reformatted for reading on screen, with the figures that are directly comparable across years pulled onto a single page.

    HTML

  2. COMPUTEX-2026 COMPUTEX-2026 Public

    COMPUTEX-2026

    HTML 2

  3. ai-medical-daily ai-medical-daily Public

    An automatically curated daily digest of AI and medical-AI news, industry moves, regulation and product analysis.

    HTML

  4. lazy-html lazy-html Public

    A Claude tool that turns text, lists, files or a topic into a polished, interactive web page — no coding required.

    HTML

  5. med-vc med-vc Public

    A global directory of medical & biomedical venture investors — life-science VCs, pharma/medtech CVCs, crossover funds, accelerators and government programs — with multi-axis filtering, full-text se…

    Python

  6. triz-40 triz-40 Public

    A bilingual codex of TRIZ's 40 inventive principles, from ancient crafts to AI — definitions, sub-principles and cross-domain examples (437 cases, 174 sources), with flashcards and quizzes.

    HTML