Yanyi Huang
Full-stack & platform engineer · quantitative finance
CQF · expected November 2026I have spent more than a decade shipping production systems across insurance, financial services, design-cloud infrastructure, computer vision, and campus platforms. The through-line is the same: take messy requirements and data, turn them into something that can be released, observed, and operated. I am based in Hong Kong and currently studying the Certificate in Quantitative Finance while building internal systems at HKU.
Work experience
The University of Hong Kong
IT Technician · Hong Kong
- Own end-to-end delivery of internal institutional systems: planning with business stakeholders, implementation, release, production support, and SDLC documentation.
- Build full-stack applications with Next.js, React, Node.js, and Nest.js for internal processes; keep the stack consistent with React Query, Tailwind CSS, and shared API-integration patterns.
- Design authentication and authorization around Keycloak: OAuth 2.0, role mapping, RBAC, and centralized identity across web apps.
- Maintain Sitecore / C# campus sites and backend customizations; also work on Drupal / PHP CMS for content operations.
- Analyse source systems and documents to identify required fields, then verify outputs against originals for accuracy and lineage.
- Use Linux-oriented tooling and shell scripts for environment checks, automation, and incident support. Cursor is part of the workflow; code review still owns quality.
FWD
Technical Lead & Full Stack Engineer · Hong Kong
- Led design and hands-on delivery of a production insurance sales platform (React, Next.js, Node.js, Nest.js, AWS, MySQL) built for sustained traffic and continuous operations with minimal downtime.
- Led and mentored a 4–10 person full-stack group across Vietnam, Thailand, Singapore, and Malaysia: coding standards, reviews, technical documentation, and delivery planning. Coaching work coincided with about a 35% lift in team productivity.
- Owned CMS and API layers that expose validated insurance product data to internal consumers. Evaluated and migrated vendor CMS from Sitecore to Contentstack, including custom APIs, data-migration tooling, and regional cutover with vendors.
- Built Python crawlers that extract insurance product data from online sources, normalize fields into structured records, and persist them for downstream analytics and model training. Reviewed extraction output against source pages for missing fields and logic gaps.
- Built lead-generation flows that wrote qualified leads into Salesforce, plus home/product-page work aimed at conversion. Segmentation and analysis supported about a 25% increase in qualified leads.
- Ran Python pipelines on AWS for claims and policy data; shipped Grafana observability and API monitoring for incident response under tight regional release cycles.
GLP
Full Stack Software Engineer & Lead · Shanghai
- Led three full-stack engineers while architecting a React + Flask + MySQL platform for monitor-area configuration and inference review. Client canvas work used Konva.js; the server shipped RBAC, document generation, CORS, and XSS/CSRF protections.
- Built an edge computer-vision / OCR stack on Jetson Nano (Python, CUDA, camera feeds) for license-plate recognition and vehicle-entry detection with PyTorch, YOLOv5, and R-CNN.
- Physical layer: perspective straighten, denoise, and contrast enhancement so skewed camera frames became upright, normalized images before OCR.
- Visual and semantic layers: detect plate regions, run OCR to structured fields with confidence scores; for unreadable characters, crop and upscale the local patch for a second pass.
- Business and HITL layers: plate-format rules and confidence thresholds; route failures to operator review on the React/Flask site; log defects and feed corrections back into retraining. Also owned dirty-data cleanup, relabeling, and dataset-quality checks before model training.
- Solar PV and energy-storage study: combined weather, historical peak demand, and sunshine duration; used NumPy/SciPy for statistical analysis and curve fitting to model generation and consumption and to size storage against ROI and grid stability.
Autodesk
Senior Software Engineer II · Shanghai
- Owned workstreams migrating an offline, GPU-intensive design-file rendering platform to AWS (Linux workers, Docker/Kubernetes): right-sizing, auto scale-up/down, production rollout, and ongoing support.
- Contributed to map-reduce-style orchestration—split jobs, run in parallel, merge results—covering IO-heavy file handling, CPU processing, and GPU image rendering, with reconciliation against source design packages.
- Worked across the stack: Play Framework (Scala) REST APIs on the JVM, Node.js scheduling, C++ rendering workers, and AWS SQS for asynchronous handoff. Spring Cloud services parsed cloud configuration and machine inventory into AWS provision/update commands.
- Led CI/CD migration from TeamCity to Jenkins (and Spinnaker on the delivery path). Rewrote build/release automation in Python, PowerShell, Ruby, and shell. Moved pipeline and service credentials into AWS Secrets Manager so secrets were not embedded in scripts or images.
- Replaced static JSON with Jsonnet to generate Kubernetes manifests and environment overlays. Implemented blue-green cutover on Kubernetes after the candidate passed pipeline gates: static scanning, functional/integration verification, smoke, preburn, and acceptance tests.
- Rewrote a ~8,000-line legacy Ruby + Windows BAT rendering pipeline into maintainable Python recipes that could be tested, replayed, and adapted for new file types.
- Built Python feature-extraction over rendered design imagery (visual attributes, layout cues, quality signals) for automated validation. Delivered ASP.NET/C# upload workflows with backend validation, C++ render invocation, database persistence, and frontend presentation.
- Designed L4+L7 traffic and API Gateway patterns for high-concurrency ingress. Production monitoring used InfluxDB, Python metric scripts, and Grafana. Participated in on-call; optimized SQL via indexing and refactoring.
ZhongAn Online P&C Insurance
Full Stack Engineer & Tech Lead · Shanghai
- Led a six-person delivery group (four developers, designer, QA) in an insurance / FinTech setting: SDLC, documentation, and production support for analytics platforms.
- Built a browser SDK that monitors DOM-tree mutations—not screenshots—to capture behaviour events. Streams were processed online for near-real-time overlays and offline in batch for cleansing, warehouse ingestion, and reporting.
- Rendered click-heat and mouse-path overlays back onto the original page by DOM coordinates, so reviewers could see behaviour in place rather than on a detached screenshot.
- Built event-driven pipelines: Kafka → JStorm transformation → HBase / MongoDB, with Hive / HDFS warehouse paths and SQL-oriented reporting. Online vs offline datasets were reconciled to keep fields accurate and replayable.
- Evaluated and integrated Alibaba DataX for sync among MySQL, HDFS, and MongoDB. Delivered an ETL job-management UI (React, Webpack, Dva.js, Redux-saga) for operational control and visualization.
Datayes
Software Developer · Shanghai
- Built a Python-centric quantitative backtesting platform: users submit strategy code in the browser; the backend parses, validates, and dispatches it to the trading engine, with explicit error handling for invalid inputs.
- Executed strategies against historical market data and streamed results—returns, drawdown, volatility, delta, Sharpe, and related risk metrics—to the frontend over WebSocket for live chart updates.
- Talked to C++ market/execution services with a custom message-framing protocol so real-time data stayed ordered and recoverable. Kept extracted and computed fields traceable to the input strategy and market feed.
- Shipped an asset-portfolio product: users select instruments, set weights, and receive historical returns, risk metrics, and optimization suggestions from the same backtest stack. Report modules ingested Excel/CSV outputs and reconciled numbers against source trades.
Shanghai Lujiazui International Financial Asset Exchange
Software Developer · Shanghai
- Worked on a financial asset-exchange platform with a microservices layout of 20+ services, including a React web app (
my-web) and Java services (my-app,asset-app) behind Spring MVC, with CDN for static assets. - Implemented build automation with Gulp, Webpack, and Gradle, and a Git workflow that distinguished quick, emergency, and normal releases, including tagging and QA coordination.
- Collaborated with the mobile team via jsbridge to ship hybrid applications on the same backend contracts.
Dell
Software Analyst · Shanghai
- Developed PHP applications with JavaScript, MySQL, OpenText CMS, CodeIgniter, and Drupal for enterprise web properties.
- Built a Python integration between RedDot CMS and the online marketing stack plus Salesforce: scheduled campaigns that matched user queries to new product content, notified sales when prospects showed interest, and stored qualified leads.
- Implemented digital-marketing instrumentation with TMS, Salesforce, and Google Analytics; ran A/B tests to tune pages against measured conversion rather than opinion.
Tech stack
Education
- Certificate in Quantitative Finance (CQF), CQF Institute — in progress, expected November 2026. Coverage of quantitative trading, derivatives, and risk.
- M.S. Computer Science, Fudan University, Shanghai · Sep 2008 — Jan 2012.
- B.S. Mathematics, Shandong Technology and Business University · Sep 2004 — Jul 2008.
Languages
- Mandarin — native
- Cantonese — native
- English — professional working
Outside of work
- Mountain hiking and long walks in nature
- Tang and Song dynasty architecture—calm lines, wood, and proportion
- Piano and guitar, mostly for unwinding
- Hiking trails and quiet meditation to reset
Contact
- hyyfrank@gmail.com
- GitHub
- Hong Kong