My Projects

AI-powered screen guidance assistant that watches your screen and provides step-by-step instructions for any software. Features include automatic screen capture and analysis, AI vision with GPT-4o, task decomposition into atomic steps, automatic progress tracking, and an always-on-top sidebar overlay.

Cross-border financial platform for instant remittances, conditional escrow payments, and real-world asset tokenization built on the XRP Ledger. Features instant transfers settling in 3-5 seconds, fees under $0.01, RLUSD stablecoin support, on-chain identity verification, and an RWA marketplace.

3rd runner-up (Top 4 out of 400+ teams), best team for Problem Statement 3: Redefining Level 2 Product Ops. Developed an AI diagnostic assistant for PSA duty officers to automates alert parsing, root cause analysis, and resolution planning with integrated ticketing and analysis system.

AI wellness companion for seniors featuring real-time voice chat, a hobby hub with 30+ activities, and the SeniorGram social feed. Accessibility-first design with multilingual support, high contrast UI, and HeyGen streaming avatars powering natural conversations and community class bookings across Singapore.

In a team of 5, we built SentinelAI, an AI-powered safety system using CCTV/IP cameras to detect falls, strokes, and industrial accidents in real time. The idea was to leverage existing camera infrast...

Finalist (Top 10 out of 60+ teams), best usage of data, best team for Theme 1: Health and Wellbeing. React Native app that integrates real-time Singapore health data, GPT-powered guidance, and community reporting to build an AI-driven, multilingual public health ecosystem.

Finalist (top 20 of 80+ teams). React Native app that uses AI and community insights to detect online scams, analyze suspicious URLs, and educate users through interactive modules, analytics, and a scam discussion forum.

Modern Next.js web app that helps users track their recovery journey, earn achievements, and stay motivated with streaks, rewards, and community support.

Top 3 (Called for top 3 in University Category out of 93 teams, final ranking undetermined due to absence from oral presentation). Developed an end-to-end NLP pipeline leveraging DistilBERT to classify vessel deficiency severity from PSC inspection reports, including data cleaning, label resolution, and model fine-tuning.