Yupo (Jason) Niu
17 years of solving hard problems. Now I solve them at AI speed.
AI Engineer in Calgary — I combine 17 years of production engineering with hands-on agentic AI system building to ship multi-agent LLM platforms that work in production.
From enterprise platforms handling 10K concurrent users to multi-agent LLM systems scoring 0.94 on quality evaluation — I've built production systems across fintech, e-commerce, smart buildings, and AI. My daily workflow integrates Claude Code, MCP, and custom AI tools to ship faster and solve harder problems.
How I Think
I don't avoid hard problems — I run toward them. Here's the methodology I've built over 17 years.
Yupo (Jason) Niu
AI Engineer · Calgary, Canada
Who I am: I'm Jason — 17 years building production systems, including 8 years designing decision engines, risk models, and workflow automation at Edianyun (HKEX-listed). Before LLMs, I built rule-based systems that automated pricing, risk, and approvals. Now I build the same class of systems with LLMs, RAG, and agents. 4.0 GPA in Data Analytics and Integrated AI at SAIT.
What drives me: I'm genuinely curious about how things work — and how they could work better. I don't just use AI to code faster. I use it to learn faster, think differently, and explore problems I couldn't have tackled alone.
How I Build
The methodology above is powered by a custom-built development environment. I don't just use AI tools — I build them.
Voice
Speak ideas naturally in any language
AI Agent
Claude Code orchestrates implementation
Code
Type-safe, tested, production-ready
Ship
CI/CD to production in minutes
VIBE Toolkit
Custom-built tools powering voice-integrated development
vox.sh
Voice-to-clipboard pipeline: Groq Whisper ASR → Gemini AI correction → instant IDE paste. Supports bilingual Chinese-English with programming terminology.
Bash · Groq · Gemini
akm
AES-256-GCM encrypted API key manager built in Go. Secure credential storage via macOS Keychain + SQLite for multi-service development.
Go · AES-256 · Keychain
claude-notify
Real-time Telegram notifications for Claude Code task completion. LLM-powered session summarization via Groq for async monitoring.
Bash · Telegram · Groq
84
Active Repos
past year
1,139+
Commits
past year
25
Projects
in development
9
Languages
in production
Experience
Every role started with a hard problem. Here's what I built to solve them.
Build a smart building management system from scratch — BMS, mobile app, AI document search, and face recognition access control — for a Canadian startup
- •Developed enterprise BMS as a monorepo: ASP.NET Core 8 backend, Next.js 15 admin panel, React Native mobile app with multi-tenant management and real-time SignalR communication
- •Built agentic AI workflow platform with LangChain, featuring intent recognition, multi-agent orchestration, pgvector hybrid retrieval, and OCR document processing
- •Developed cross-platform mobile app using React Native + TypeScript + Expo with Firebase auth and biometric features
- •Implemented CI/CD automation using GitHub Actions with Docker + GCP deployment
Monorepo serving admin, mobile, and AI services. Shipped to GCP with CI/CD.
AI Developer
Build and scale enterprise decision systems for a publicly traded company — financial algorithm engines, real-time risk control, workflow automation, and distributed architectures handling 10K+ concurrent users across 8 product platforms
- •Built core financial algorithm engine implementing NPV, ACPI, and ROC calculations — the pricing decision engine powering all rental orders, buyout pricing, and risk reserve computations across multiple leasing models
- •Designed real-time risk control system with Alibaba DTS data pipelines + Kafka streaming for multi-dimensional credit scoring, automated risk alerting, and real-time indicator monitoring
- •Architected decision automation platform with multi-level approval workflows (M1→M2→M3→city manager), quota pool management, and 10+ configurable coupon/incentive policy types integrated with HR position hierarchy
- •Led 3 development teams (Risk Control, E-commerce, DevOps) across 8 product platforms, responsible for technical design, code review, and project delivery
- •Built custom Dubbo RPC framework and Spring Boot Starter library as shared infrastructure for all microservices, plus unified RBAC permission center serving 1K+ users across business units
- •Migrated monolith→Spring Cloud microservices→Kubernetes with Jenkins/GitLab CI/CD and blue-green deployments, built observability stack (ELK + Prometheus + Skywalking APM) — significantly reduced API response times
Led 3 teams, built 8 product platforms including financial algorithm engines, risk control systems, and workflow automation. Before LLMs, I built rule-based systems that automated pricing, risk, and approvals. Now I build ML-driven systems solving the same business problems with LLMs, RAG, and agents.
Senior Full Stack Developer / Team Lead
Build financial systems for the world's largest crypto mining company — mining pool revenue distribution, a Bitcoin payment app, and automated trading
- •Maintained and optimized Antpool mining pool system, developed PoW revenue distribution algorithm with high-concurrency hashrate statistics
- •Independently developed internal Bitcoin payment application enabling employees to use Bitcoin for daily purchases
- •Integrated with major exchange APIs (Huobi, Binance, OKEx) for automated arbitrage trading
- •Participated in wallet API design, implemented Bitcoin transaction confirmation and blockchain data synchronization
Independently shipped Bitcoin payment system, optimized blockchain confirmation time 40%. Early exposure to distributed computing and real-time data patterns now essential for AI infrastructure.
Full Stack Developer
Deepen AI specialization — computer vision, predictive modeling, AI governance, and applied AI project delivery
- •Core courses: Computer Vision, Predictive Analytics & Modeling, AI Governance & Ethics, Human-Centred AI, AI Management & Maintenance
- •Capstone: Applied AI Projects (PROJ-407-A) — production-grade AI system development
- •Graduated April 2026 with 3.92 program GPA (7 A+ across 8 courses)
Completed — 3.92 GPA. Combined with Data Analytics, total 16 courses, 15/16 A+, overall 3.96 GPA across 54 credits
Post-Diploma Certificate in Integrated Artificial Intelligence
Build data foundations — statistical analysis, predictive analytics, and business intelligence to complement 17 years of engineering experience
- •Achieved 4.0/4.0 GPA while working part-time on real-world projects with local development teams
- •Core courses: Statistical Analysis, Predictive Analytics, Business Intelligence Reporting, Business Analytics
- •Capstone: cancer recurrence prediction system achieving 97.4% accuracy with XGBoost, SHAP interpretability, and Streamlit dashboard
4.0 GPA, capstone ML system (97.4% accuracy), solid foundation for AI specialization
Post-Diploma Certificate in Data Analytics
Projects
Every project started with a question I couldn't leave alone.
Can an AI agent find, score, build resumes for, and apply to jobs autonomously?
JobPilot AI — Agentic Job Automation
AI & Machine LearningAutonomous multi-agent job search pipeline orchestrating 5 LLM providers with circuit breakers, structured output, reflexive quality gates, and auto-apply via real Chrome. 1,250+ tests, 4 discovery sources, Telegram bot with inline keyboards.
How do you make enterprise documents searchable without structured metadata?
Industry-AI-Flow - AI Workflow Platform
AI & Machine LearningEnterprise-grade AI workflow system with intelligent intent recognition, routing, and multi-agent orchestration. Features hybrid retrieval (BM25 + vector search) with pgvector, OCR document processing, and code execution capabilities.
AI Ops Control Room - LLM Quality Evaluation
AI & MLOpsLLM quality automation system using LLM-as-Judge pattern. Local Qwen 3.5 simulates e-commerce customer service, DeepSeek V3 evaluates responses on relevance, faithfulness, and effectiveness — achieving 0.94/1.00 composite score.
Can local LLMs make profitable trading decisions in real-time?
Trading Bots - LLM-Driven Automated Trading
AI & FinTechMulti-user SaaS trading system where local LLMs (Qwen 2.5, DeepSeek-R1) make entry/exit decisions. TradingView webhooks trigger signals, AI evaluates market context via customizable prompts, and trades execute through IBKR API with per-user/per-symbol configuration.
Can AI hand tracking be fast and accurate enough for real-time gaming?
Fruit Ninja AI — Hand Gesture Game
AI & Computer VisionGesture-controlled fruit cutting game using real-time AI hand tracking via MediaPipe. Three.js 3D rendering with fluorescent neon trails, particle effects, and adaptive performance. Supports webcam + mouse/touch fallback, deployed on Alibaba Cloud ESA.
Can one team build a BMS, mobile app, and AI system in a single monorepo?
HavenzHub - BMS & Mobile Platform
Full StackEnterprise building management system with mobile app, AI-powered document RAG, and face recognition access control. Monorepo architecture with ASP.NET Core backend, Next.js admin panel, React Native mobile, and Python AI services.
More Projects
Vox - AI Voice-to-Text Pipeline
AI & ToolsDual-endpoint (CLI + iPhone Web) voice-to-text tool with a full ASR → AI correction → pronoun verification pipeline. Uses Groq Whisper for transcription with Gemini 2.5 Flash for intelligent text correction, served via Cloudflare Tunnel.
AKM - AI API Key Manager
AI & DevOpsLightweight daemon-mode API key vault for AI platforms. Secures keys with AES-256-GCM encryption and macOS Keychain integration, serves them to local scripts via Unix socket API. Supports 7 AI providers (OpenAI, Anthropic, Gemini, Groq, DeepSeek, Qwen, GLM).
Local Chat RAG - Privacy-First AI Chat
AI & Machine LearningPrivacy-first Retrieval-Augmented Generation chat application. Upload documents, ask questions, and get answers with source citations — everything runs locally on your machine using Ollama.
AgenticAI2026 — AI Agent Curriculum
AI & Machine LearningInteractive AI Agent learning curriculum for senior developers. Covers multi-agent patterns, tool orchestration, MCP, and production deployment of agentic AI systems.
HockeyAI-Tracker - Real-Time Player Tracking
Computer VisionReal-time hockey player and puck tracking system using YOLOv8 object detection with BoT-SORT multi-object tracking. Generates CSV statistics and annotated video output with team-based tracking.
Talk2Type - Smart Voice-to-Text for macOS
Native AppNative macOS voice-to-text dictation tool with global hotkey activation. Supports 8 ASR services (OpenAI Whisper, Groq, Alibaba, Tencent, Baidu, iFlytek, AssemblyAI, Speechmatics) with cost optimization and multi-language localization.
Golf Swing Analyzer - Biomechanical Analysis
Computer VisioniOS app for real-time golf swing biomechanical analysis using Apple Vision framework. Detects 7 biomechanical metrics and 11 issue types at 30 FPS with pose estimation, providing scoring (0-100) and video import support.
YiPaiJi - IT Equipment Auction Platform
Full StackHigh-concurrency auction platform handling thousands of simultaneous bidders. Optimized with WebSocket real-time updates, Redis caching, and RocketMQ asynchronous processing.
ATV-Bilibili - Apple TV Streaming Client
Native AppFeature-rich BiliBili streaming client for Apple TV (tvOS). Extensively customized fork with QR login, live streaming with real-time comments, HDR/Dolby Vision support, SponsorBlock integration, and playlist management.
What I'm Exploring
I design custom AI learning curricula to keep up with a world that changes every week.
I track GitHub trending, follow AI research on X, and build custom learning paths to rapidly acquire domain knowledge in areas I haven't mastered yet. These are the threads I'm currently pulling on:
Skills
From AI systems to production engineering — 17 years of solving hard problems.
AI & LLM Systems
Machine Learning & Data
Production Engineering
AI-Augmented Development
Languages
Get in Touch
I'm looking for teams where the problems are hard and the standards are high. If that sounds like yours, let's talk.