Technical Project Manager · AI Automation and Copilot Delivery
I'm a Technical Project Manager at TELUS Digital with an AI/ML engineering and backend development background. I run AI and automation projects end-to-end, from writing the business case and running the pilot to UAT and handover, and I can build the technical solution myself. My recent work includes AI copilots, KPI dashboards, OCR pipelines, and cross-system automations using Fuel iX, Workato, and Google Apps Script.
Technical Project Manager
TELUS Digital
Aug 2026 - Present
Operation Team Leader
TELUS Digital
Jan 2023 - Jul 2026
Backend AI Engineer - Intern
FlyRank AI
July 2026 - September 2026
Backend Developer (Freelance)
Cost-Effective Software
Dec 2024 - Mar 2025
Computer Science, BSc
University of London
IBM
Workato
Leadership & Process
Other
Project Management
Tools
AI, ML & Data
Backend & Infrastructure
Languages
AI-powered KPI dashboard built with Fuel iX and Google Apps Script, serving 10 team leads, ~170 agents, and 16 accounts. Includes an AI chat assistant with per-account KPI rules. Projected savings: ~83 hours/month.
Fuel iX copilot that auto-generates test cases for the HR team, cutting test preparation time from hours down to ~5 minutes.
Suite of Fuel iX AI copilots deployed across multiple operations projects to reduce manual work and support agents in real-time, improving throughput and consistency.
Automated change-ticket documentation for 149 accounts using Google Apps Script, the ServiceNow API, and Fuel iX AI, with a change log in Google Sheets and automatic email confirmations to account managers.
OCR pipelines using PyTesseract and EasyOCR combined with a fine-tuned Donut transformer model for structured document data extraction and understanding.
Automation workflows built in Workato to connect and synchronize data across different enterprise systems, reducing manual handoffs and improving operational efficiency.
Semantic image-to-blog-post matching system built for FlyRank AI. Uses Google Gemini for vision captioning and embeddings, cosine similarity for ranking, and a mismatch guard that rejects low-confidence suggestions. Achieved 75% Top-1 Precision on a labeled evaluation set.