Qijin Xu

Summary

Lead AI/ML Architect and Tech Lead at CGI’s AI Center of Excellence with 5+ years delivering LLM agents, RAG platforms, and ML systems to enterprise production for major banks and insurers. Two-time CGI Impact Award winner spanning agent architecture, evaluation, technical leadership, and multi-cloud AI delivery. Built an LLM-powered RAG assistant that resolved about 90% of questions on its own for 50,000+ users and returned 525% ROI within 12 months. Now leading a LangGraph agent for a proprietary programming language, projected to cut manual work by 50%.

Experience

1. 2025

   Lead AI/ML Architect & Tech Lead, AI Center of Excellence

   CGI Inc. · Toronto, ON · Jan 2025 – present

   • CGI Impact Award 2026: led development of an LLM-powered email agent for enterprise email workflows. 89% autonomous handling across 2,000+ emails per month.
   • Lead an AI agent for a proprietary programming language: LangGraph dual-state dynamic workflow, vLLM-served LLMs with RAG, EBNF structured outputs, and PII filtering. It writes documentation, extracts business rules, maintains legacy scripts, and prepares them for migration. Projected 50% less manual work.
   • Architected a scalable agent evaluation pipeline with SME-adjudicated rubrics, reference cases, regression suites, and calibrated LLM-as-a-judge evaluators, for decision tasks with noisy and conflicting human labels.
   • Drive technical direction across 13+ AI initiatives for TD Bank’s compliance SBU, guiding architecture and delivery across a cross-functional team of about 20.
2. 2023

   Data Architect & Engineering Lead, Senior Consultant

   CGI Inc. · Toronto, ON · Aug 2023 – Dec 2024

   • CGI Impact Award 2024: architected and shipped an LLM-powered enterprise RAG assistant for insurance support (LangChain; Azure, AWS, GCP). About 90% autonomous resolution for a 50,000+ user base; 525% ROI within 12 months. Product page: cgi.com/canada/en-ca/insurance-chatbot
   • Built an OpenLineage-based lineage tool that auto-documents multi-source data flows, with LLM-assisted root-cause suggestions. Reduced investigation time across about 3 incidents per month.
   • Automated AxiomSL rule updates at RBC with an LLM: 80% less manual updating, governance controls preserved.
3. 2021

   Data Engineer, Consultant

   CGI Inc. · Toronto, ON · Aug 2021 – Jul 2023

   • Managed Canada’s largest automobile-insurance policy-history database (2 TB, SQL Server): 99.9% uptime, median queries under 200 ms.
   • Built production ETL/ELT pipelines for 22+ micro data marts (under 6 tables each, monthly SLAs) for finance, analytics, and delivery teams. Report runtime cut from 30 to 5 minutes.
   • Shipped a gradient-boosting and logistic-regression loss-ratio model, batch-scored in a Python/T-SQL pipeline: 321% lift in recovered fraud losses versus a top-five bank benchmark.
4. 2020

   Research Assistant, Computer Vision (part-time)

   Elder Laboratory, York University · Toronto, ON · May 2020 – Aug 2021

   • Worked on AttentiveVision’s computer-vision pipeline for hockey broadcast video (PyTorch, TensorFlow, OpenCV). Trained a CNN-LSTM stack reaching mAP ≥ 70% on event detection across 18 hours of game video.
   • Built real-time player tracking with a Kalman filter and OpenPose: under 100 ms latency at 30 FPS, for live speed and heat-map analytics.
   • Built an event-detection annotation loop (goals, timeouts, and penalties) with active learning, raising label reliability to Cohen’s kappa ≥ 0.92.
5. 2019

   Full-Stack Developer Intern

   Nascent Digital · May 2019 – Aug 2019

   • Built features in React, TypeScript, and Node.js, including major parts of the company homepage rewrite, and shipped TELUS product pages with UX, QA, and DevOps, design to deploy in 7 days.
   • Introduced visual-regression testing (Nightwatch.js, Puppeteer), removing about 90% of manual QA checks.

Skills

Languages

Python, Java, Go, SQL (T-SQL, PostgreSQL), TypeScript/JavaScript, C++

AI/ML

PyTorch, JAX, LangGraph, RAG, agent workflows, Model Context Protocol (MCP), LLMs, vLLM, OpenAI API, FAISS, Pinecone

Data and streaming

Apache Spark, Kafka, Airflow, OpenLineage, Redis

Cloud and systems

Azure, GCP, AWS, Docker, Kubernetes, Terraform, microservices, gRPC, CI/CD

Education

York University, Toronto, ON. B.Sc. Computer Science, Specialized Honours, 2017 – 2020.

GPA 3.9/4.0 · First in Class · Dean’s List (all semesters)

Awards and certifications

• CGI Impact Award (2024), for the enterprise RAG assistant
• CGI Impact Award (2026), for the email agent
• Microsoft Certified: Azure Solutions Architect Expert (2022)
• Google Cloud Professional Cloud Developer (2023)