William Leung

• Architected an enterprise RAG platform on AWS for ESG data processing[cite: 18].
• Shipped AI agents that automated 85% of manual ESG workflows, cutting report generation time by 90% (from weeks to days)[cite: 19].

• Engineered a fully AWS S3 compliant API (GetObjectAttributes) in C++ for the ONTAP storage system, supporting 25K+ concurrent requests for 10K enterprise clients[cite: 23].
• Delivered production-ready code with 97% test coverage across 70+ tests, achieving a zero-defect release[cite: 24].

Ensured fair and consistent grading for a 400+ person class by coordinating a team of 14 graders[cite: 27].

• Led a cross-functional team of 7 engineers through Agile sprints, delivering 9 major feature releases on schedule[cite: 32].
• Delivered a real-time fleet management app tracking 2K+ vehicles with 99.9% uptime[cite: 30].
• Implemented a CI/CD pipeline, accelerating feature delivery by 20%[cite: 31].

• Enhanced an LLM reasoning framework by integrating a custom reinforcement learning component in PyTorch.
• Reduced API costs by 100x and latency by 75% while maintaining 92% accuracy on math reasoning tasks[cite: 36].

• Built an end-to-end ML pipeline, scraping 30K+ fashion items from 7 e-commerce sites[cite: 38].
• Fine-tuned a sentence transformer, allowing users to search with natural language to get outfit recommendations[cite: 38].

• Deployed a computer vision pipeline processing 70K+ BirdCam frames daily using YOLO[cite: 41].
• Built ML models to predict harmful algal blooms with 85% accuracy[cite: 42].
• Created a visualization that identified 15% potential energy savings on campus[cite: 43].