IGC — Digital Twin & AI Integration

Built 6 RAG (Retrieval-Augmented Generation) knowledge bases from 1,000+ building documents covering MVAC, Electrical, Plumbing, Fire Services, Lift systems, and more. The pain point: building maintenance requires specialized engineers for even minor issues, creating slow response times. The solution: an AI Agent that can answer questions by referencing the actual documentation.

My role was to organize and upload these documents into the AI system. I wrote Python scripts to scan the 1,000+ files, identify their category by filename, organize them into the correct folders, then batch-upload them to the agent. I tested the agent by crafting domain-specific questions to verify it could accurately locate the right documents and cite them correctly.

Contributed to API design for the Henderson Digital Twin project. For each feature update, I needed to understand the purpose, identify which API endpoints to use, determine polling frequency, decide how many data points to display, and specify the UI placement — essentially acting as a product manager.

Contributed to a smart transformation proposal for HK Housing Authority, researching technologies like NFC access control, fall detection sensors, overhead object detection, and parking space monitoring. The proposal centered on Building OS — a unified interface integrating all building management data, resident management, predictive maintenance, and energy optimization.

When IGC needed compelling ESG data for an ICT Awards application but had no historical baseline (the system was just launched), I designed a 'horizontal benchmarking' approach: comparing theoretical energy consumption from EMSD benchmarks against actual electricity bills, adjusted for real occupancy rates (73%). This model was iteratively refined through consultation with senior PMs and successfully defended to both internal teams and clients.