🟢 Open to work — PGWP work-authorized, no employer sponsorship required. Available immediately for permanent full-time roles in Calgary or remote Canada.
📬 Reach me: yupoca24@gmail.com · LinkedIn · portfolio.dissidia.tech · Resume (PDF)
17 years of production engineering. 8 years as Team Lead at a publicly traded company designing decision automation systems — financial pricing engines, real-time risk scoring, approval workflow platforms. Now applying that same foundation to LLM and agentic systems: multi-agent orchestration with provider fallback chains, enterprise RAG, LLM evaluation pipelines.
Recently graduated SAIT (April 2026): Post-Diploma Certificates in Integrated AI (3.92 GPA) and Data Analytics (4.0 GPA), 15 of 16 courses A+.
Python FastAPI LangChain LangGraph pgvector PyTorch XGBoost · Java Spring Cloud Kafka Kubernetes · Go Swift · React Next.js TypeScript React Native · PostgreSQL Redis · Docker GCP Vercel
| Project | What it does |
|---|---|
| go-mcp-bridge | MCP bridge that lets an LLM agent drive an existing Go ticketing service — authorization gates live in the bridge, so the agent can't exceed its permissions |
| Industry-AI-Flow | Enterprise RAG platform on LangChain — BM25 + pgvector hybrid retrieval over multi-format document corpus (PDF, DOCX, scanned via PaddleOCR) |
| AI Ops Control Room | LLM-as-Judge evaluation pipeline (Qwen 3.5 simulates customer, DeepSeek V3 evaluates on DeepEval, 3-dimension rubric) — React + FastAPI |
| Talk2Type | Native macOS dictation daemon in Swift — global hotkey, 8 ASR providers (OpenAI, Groq, DashScope, Tencent, Baidu, iFlytek, AssemblyAI, Speechmatics), transcribes and pastes straight into the active field |
| Hockey Puck Tracker | Computer vision demo — YOLO baseline + RF-DETR/RT-DETR for puck detection, multi-object tracking, trajectory visualization |
| Fruit Ninja AI | Gesture-controlled WebGL game — MediaPipe real-time hand tracking + Three.js 3D rendering on Alibaba Cloud ESA edge |
| diabetes-prediction-app | End-to-end ML pipeline for diabetes risk prediction — PyCaret + Streamlit, deployed on Streamlit Cloud |
| CELPIP-Compass | Full-stack spaced-repetition learning system — Next.js 16 + Prisma + PostgreSQL + Redis, Jest + Playwright coverage |
Live portfolio site: portfolio.dissidia.tech (Next.js 16 + React 19 + Tailwind v4, deployed on Vercel)
I treat AI features the same way I treat any production system: explicit interfaces, reliability stacks (circuit breakers, retries, structured output), observability, and tests.
I keep a few production projects private (multi-tenant SaaS work, prior Havenz Tech codebase) — happy to walk through architecture and code in interviews.
JobPilot AI is one of those — an autonomous multi-agent job pipeline orchestrating 5 LLM providers (GLM, Qwen, DeepSeek, GROQ, Ollama) with circuit breakers, structured Pydantic output, and per-task fallback chains. Live demo at jobpilot.dissidia.me (login required, self-hosted on Cloudflare).
If you're hiring for AI / LLM Engineer, Senior Full-Stack with AI scope, or AI Platform / MLOps roles in Calgary or remote Canada — let's talk.