AI Governance Leadership · Agentic AI Engineering · AWS Solutions Architect Professional
18+ years of cross-border engineering experience (Germany + Hong Kong) — ERP systems, applied AI engineering, and scalable cloud solutions. PMP + CISA certified; working toward recognized cybersecurity and AI safety standards.
Currently Head of AI Governance at Votee AI (Hong Kong), where I drive our cybersecurity and AI safety standards programme toward certification alongside the AI governance function.
📄 Full portfolio, live projects, and case studies: carsonng.com
- Command Deck — ops center for a fleet of live services: cross-project LLM cost tracking, health monitoring, and a deliberate quarantine/restart-eligibility gate that distinguishes an intentional pause from a genuine liveness failure. · Live · GitHub
- Quantitative Trade-Analysis Platform — a 22-ETF trading platform where a multi-agent LLM analyst feeds a deterministic risk gate that decides whether capital moves, not the LLM itself. Live on Interactive Brokers alongside an isolated paper account. · Paper Account · GitHub
- AWS AI Code Review & Security Scanning Demo — Amazon Q Developer + Amazon Inspector gating merges on critical findings, not leaving them as comments someone might read. · GitHub
- Event Radar — AI event discovery for Hong Kong with a closed feedback loop and hallucinated-event-ID rejection on every LLM-referenced result. · Live · GitHub
- SpendLens — personal credit-card spend monitor where pandas computes every figure from the source statement, so the model structurally cannot invent a number. · Live · GitHub
- AI Regulation Radar — monitors EU AI Act, NIST AI RMF, and HK PCPD sources, diffs legal text changes, and generates plain-English impact assessments. · Live
- Sprint Analyzer — deterministic-numbers / LLM-prose split: pandas computes every metric from Jira/ClickUp exports, the LLM only writes prose. 110+ tests run without an API key. · Live · GitHub
- Change Impact Assessor — a risk assessor for code/infra changes where a HIGH-risk classification actually pauses the workflow behind a real human-in-the-loop checkpoint, not just an advisory label. · Live · GitHub
PMP · CISA · AWS Certified Solutions Architect – Professional · HashiCorp Terraform Associate · PSM II · Microsoft Azure AI Engineer Associate · Zend Certified Engineer
AI/ML: LangGraph, LlamaIndex, RAG, multi-provider LLM orchestration (Anthropic, OpenAI-compatible, DeepSeek), Supabase/pgvector Security & Governance: AI governance · AI safety · cybersecurity & information-security management · NIST AI RMF · EU AI Act · human-in-the-loop design · deterministic risk gates Cloud & Backend: AWS, Terraform, Docker, PHP/Laravel, Node.js, Python, GitHub Actions Frontend: React, React Native, Vue.js
📫 Reach me @ info@carsonng.com · 💼 linkedin.com/in/carsonnghk