← Back

Accenture Innovation Hub

Software Engineer, Natural Resources
Oct 2018 – Jan 2020 · Singapore

Built 10+ rapid prototypes to pitch emerging tech to industrial clients across energy, mining, agriculture, and transportation. Half R&D, half sales engineering: show what's possible, hand it to delivery.

Highlights

Distributed Energy Marketplace

Simulated peer-to-peer trading of tokenized distributed energy assets on a blockchain-backed market. Built to make a case to Southeast Asian utilities and regulators that deregulation is opportunity, not threat.

Stack: Hyperledger Fabric, Node, AngularJS, Docker, Postgres

ENC.iQ

A multilingual voice agent. ("Encik" is honorific Malay for "uncle"; the acronym stands for Enabling Natural Communication.)

ENC.iQ started as a chatbot that auto-adjusted to the speaker's language (English, Bahasa Melayu, or Thai), then connected to local IoT systems to act on what it heard. Agentic industrial interaction, years before ChatGPT and Anthropic made the voice-to-MCP pattern mainstream.

Stack: DialogFlow, Python

Intelligent Operations Center

The Hub's signature demo platform: industrial IoT, predictive analytics, and natural interfaces, woven into a single control center.

A robotic arm responded to LEAP Motion hand gestures. YOLO-based computer vision (see PPE Detection) fed the dashboard. ENC.iQ took voice commands at the front. The whole thing demonstrated where industrial UIs could go once we stopped assuming the operator could only press buttons.

Stack: React, Node/ExpressJS, Firestore, LSTM-based predictive analytics, LEAP Motion SDK, YOLO, PyTorch

PPE Detection

Computer vision that flagged workers missing required PPE or doing something unsafe on hazardous industrial sites.

One prototype, two models. I trained YOLOv3 on a small set of client-supplied images to detect PPE, and an LSTM classifier watched for unsafe activity: falls, walking into danger zones, smoking. With minimal training data, YOLO made a great classifier, so we reused it across many Hub demos, including the Intelligent Operations Center dashboard. The client liked the demo, and an Accenture delivery team went on to deploy it at several customer sites.

Stack: Python, PyTorch, YOLOv3, LSTM

Name Card Reader

Point a phone at a business card and get a new contact. OCR pulled the name, title, phone, and email off the card and filled in the phone's contact form.

Connected Worker

An extended-reality wearable for field workers on Microsoft HoloLens. The headset enabled real-time voice and video calls back to the Intelligent Operations Center. Hands-free comms for field jobs that needed both hands on the work while receiving remote coaching or monitoring.

I scrum-mastered the build - my first experience formally leading other engineers.

Stack: Microsoft HoloLens, Unity, WebRTC