I do not know Jason yet.
Start with the Now page, then choose Research, Systems, or Writing.
研究與寫作
把複雜的高風險 AI 工作整理成可被信任的證據路徑。
這是 Jason Chia-Sheng Lin 的個人研究網站與寫作空間：整理近期在可信任 AI、語音決策穩定性、臨床流程支援、AI agent governance、資安與 AI 系統工程教學上的公開安全成果。
Recent work is summarized from public-safe repo artifacts and planning summaries. Raw planning notes, private contacts, patient-like data, raw transcripts, credentials, and patent-sensitive mechanics are intentionally kept out of the website.
可以從這裡開始
I am reviewing or evaluating.
Start from research programs, public artifacts, and validation paths.
I want to inspect systems.
Open the Systems page and inspect each workflow by capability, evidence, and next validation layer.
I want to learn from the work.
Use Teaching and Writing for long-form notes, course surfaces, and public framing.
近期重點
7-day Enterprise Voice AI Systems onboarding tutorial
Packaged the AI Systems Engineering Handbook into a 7-day consulting-style onboarding path for enterprise voice AI, AI Gateway, agent governance, red teaming, K8s, GPU sizing, and customer acceptance evidence.
CDS-ASR / JANUS evidence hardening
Advanced a speech-to-decision stability research line that tests whether downstream decisions remain stable under plausible ASR alternatives.
PB-EGP and Support-Transfer Validation
Developed provenance-bounded evidence graph packet work for small-model decision support, with release gates, audit protocols, and public-benchmark orientation.
AI Triage Kiosk demo
Built a synthetic vital-aware intake and staff-review summary demo for a June market demonstration, with a narrow API contract and explicit clinical-scope controls.
UroPrevisit Navigator
Hardened a urology previsit workflow prototype for adaptive question navigation, missing-field repair, clinician-review summaries, and PSA/CRM-ready proposal framing.
研究方向
Counterfactual Decision-Stability ASR
A speech-to-decision stability research line that asks whether downstream decisions remain stable under plausible transcript alternatives.
Provenance-Bounded Evidence Packets
Research on evidence graph packets that help small models make more stable, better-grounded decisions under the same token budget.
False Governability and Runtime Governance
A technology-governance research line on how action-capable AI can appear governable while reconstructability collapses underneath.
Security Evidence and CaseTrace
A security-evidence research direction that turns reviewer feedback into traceable public-source case evidence, uncertainty labels, and baseline comparison.