ZI-YU (DAVID) HUANG

AI Engineering Lead | From PoC to Production | LLM & Geospatial AI

Taipei, Taiwan | moon.starsky37@gmail.com | linkedin.com/in/davidhuang0304 | github.com/moonstarsky37 | Google Scholar: scholar.google.com/citations?user=RNvt85EAAAAJ

Professional Summary

AI engineering lead who takes AI systems from proof of concept to signed government contracts and production operations. Led a team of 3–4 AI engineers delivering 12 government and enterprise projects in 2024, each shipped with a formal test plan and passing client user acceptance testing (UAT) and system integration testing (SIT); converted an on-premises LLM proof of concept for the New Taipei City Fire Department into a signed 2025 contract. Co-inventor of Taiwan Invention Patent No. I925616 for TronCube, a 17-microservice geospatial data-warehouse platform with GitLab CI/CD. Reduced AI model delivery cycles by 70% (3–4 months to under 1 month) through module standardization. 6+ years of Python development, grounded in test automation and MLOps.

Core Competencies

AI / Machine Learning: PyTorch, Computer Vision (YOLO, Mask R-CNN, U-Net, Detectron2, OCR), Time-Series Forecasting (Transformer, ConvLSTM), Large Language Models (RAG, Fine-Tuning, Function Calling, Ollama, Whisper), LiDAR Point Clouds, Remote Sensing, Edge AI (NVIDIA Jetson, DeepStream)

Cloud & Infrastructure: Docker, Kubernetes, Terraform, Google Cloud Platform (GCP), AWS, Airflow, Proxmox VE, Grafana, Loki, RabbitMQ, PgBouncer, GeoServer, Infrastructure as Code, Observability, MLOps

Testing & Quality: Test Automation, Test Plan Authoring, Unit Testing (unittest, pytest), Integration Testing, Stress Testing, System Integration Testing (SIT), User Acceptance Testing (UAT), SonarQube Quality Gates, CI/CD (GitLab CI), Test Dataset Design, Annotation Standards, Defect Tracking, SLA Operations

Programming: Python (primary, 6+ years), SQL (PostgreSQL), RESTful API (FastAPI, Swagger); prior experience: Node.js, C++

Leadership: Team Management, Cross-Functional Collaboration, Matrixed Organizations, Client Delivery, Technical Training (English), Patent Drafting

Professional Experience

Independent Software Engineer (Freelance)

Taipei, Taiwan | Jun 2025 – Present

Delivered a dividend-reinvestment modeling tool end to end for a private client: requirements interviews, financial modeling (projecting long-term share accumulation and total cost basis from annual contributions, cash and stock dividends, and annual average market prices), and delivery in two forms — a Python tool and a Google Sheets version for non-technical use.
Supported pre-launch quality assurance for mosir.app, a social micro-blogging platform: built an article-posting bot against the platform's content APIs and ran integration tests on new API endpoints with bug reporting.
Team entrant in the Magical Mirai 2025 Programming Contest (Tokyo), building an interactive lyric-driven web application on the TextAlive App API.

Section Manager (AI), Remote Sensing & AI Solutions — ThinkTron Ltd.

Taipei, Taiwan (geospatial-AI joint venture of Kokusai Kogyo (KKC), Japan, and Sinotech Engineering Consultants, Taiwan) | Jan 2023 – May 2025

Led a team of 3–4 AI engineers delivering 12 government and enterprise projects in 2024 while operating 25+ geospatial data services.
Institutionalized quality engineering: every delivery shipped a formal written test plan and passed client UAT and SIT; all project code passed SonarQube quality gates; defined a six-stage R&D process covering unit, integration, and stress testing with defect tracking.
Architected TronCube, a geospatial data-warehouse product of 17 microservices with unittest/pytest suites, one-command test targets, and GitLab CI; co-inventor of Taiwan Invention Patent No. I925616; adopted by 4+ government projects.
Re-scoped the New Taipei City disaster-response cloud platform to RESTful-API system integration when contractor hardware stalled, shipping on schedule and saving ~2.3 person-months.
Unified the company's time-series AI modules, cutting model delivery from 3–4 months to under 1 month; personally built the Taipei sewer water-level forecasting model (Transformer), validated on Typhoon Gaemi (2024) and deployed to production.
Delivered an on-premises LLM proof of concept for the New Taipei City Fire Department that converted into a signed 2025 contract; rolled out ChatGPT-assisted development team-wide and authored Claude adoption guidelines.
Presented 3 technical topics at the KKC symposium in Tokyo; delivered 4+ English-language training courses for APRSAF (Asia-Pacific space agencies forum) and enterprise clients in Vietnam.

AI Algorithm Engineer — ThinkTron Ltd.

Taipei, Taiwan | Mar 2020 – Jan 2023

Built containerized image-recognition and OCR services for the National Fire Agency's 5G AI rescue platform, streaming results in real time to AR glasses and the command center; deployed on the agency cloud for search-and-rescue teams.
Developed a nationwide rainfall nowcasting system (ConvLSTM) producing 3-hour forecasts every 10 minutes, feeding flood simulation via RESTful APIs; designed per-source error logging; operated reliably for multiple years.
Automated road-feature extraction (poles, lane markings) from LiDAR point clouds with SLAM for a Ministry of Economic Affairs HD-map program; defined annotation standards covering 5 object classes with 200+ labeled samples each; techniques entered the company patent portfolio.
Developed nearly all recognition models for a subscription AI pavement-inspection service (most classes at 85%+ mAP) and integrated in-vehicle edge hardware; supported commercial operations under a 24-hour-fix SLA.
Served as project and technical lead for an underground heritage-tunnel monitoring service (stereo camera + NVIDIA Jetson, IP66 enclosures, 4G uplink, real-time deformation measurement, earthquake alerts) — signed and delivered.
Achieved 95%+ validation accuracy with a Mask R-CNN landslide-detection model; built an aerial-image labeling and expert-review toolchain for the Forestry Agency, semi-automating a fully manual interpretation workflow.

Earlier Experience

Military Service, Taiwan | Sep 2019 – Mar 2020.
AI R&D Intern — ThinkTron Ltd., Taipei | Jun 2019 – Sep 2019. Developed data-processing and visualization components; contributed to the company entry that won the Taiwan R&D Award at the 2019 Asia Open Data Challenge.
AI R&D Intern — Coretronic Corp., Hsinchu | Jun 2018 – Aug 2018. UAV aerial-image object detection; 2D image segmentation models.
Full-Stack Developer — Industry-Academia Cooperation (Export Trade System), Taipei | Jul 2016 – Jun 2017. Node.js service development, relational database design, and production deployment.

Education

National Taiwan University — M.S., Industrial Engineering | Aug 2017 – Aug 2019. Thesis (with National Taiwan University Hospital, on hospital-provided CT data): 3D medical-image segmentation using sequential fully convolutional deep-learning networks.
National Chengchi University — B.S., Mathematical Sciences; B.B.A., Information Management (double major) | Aug 2013 – Jun 2017. Top 7% of department.

Patent & Selected Publications

Taiwan Invention Patent No. I925616, "Geospatial Data Processing System and Method" (co-inventor; filed 2025, granted 2026, term to 2045); led patent drafting and filing.
Variational Grid Setting Network, IEEE IALP 2017 — first author, oral presentation (Singapore).
Two journal articles (2021) on AI for road-asset management and slope monitoring — J. CICHE 48(2); Geological Quarterly 40.

Certifications, Honors & Languages

Hands-on Google Cloud training (Kubernetes, Terraform, BigQuery, Cloud Security; official skill badges); 2 NVIDIA Deep Learning Institute certificates (Jetson Nano, DeepStream).
Finalist (captain, 6-person team), "City Dashboard" Hackathon, Taipei Coding Festival 2023; Best Paper Award, IISE Asia 2018; invited speaker, PyCon Taiwan.
Languages: Mandarin Chinese (native); English (professional working proficiency — international presentations and 4+ courses taught in English); Japanese (basic).