I build intelligent software, end to end.
CS undergrad at NUS and AI Engineer Intern at Singapore Airlines. I work across the stack - from YOLO-based computer-vision pipelines to production web apps in React and Go.
mAP@50 0.76
CV model
About
I'm Ike - a Computer Science undergraduate at the National University of Singapore, focused on AI/ML and full-stack engineering. I like problems that span the whole stack: training and deploying models, designing REST APIs, and shipping interfaces people actually enjoy using. Lately that's meant computer-vision pipelines for aircraft inspection at Singapore Airlines and a full forum platform on a Go + PostgreSQL backend. Based in Singapore.
Experience
Where I've worked
AI Engineer Intern · Singapore Airlines
May 2026 – PresentSingapore
Built a YOLOv8 pipeline to detect paint-peel defects on aircraft fuselages, using SAHI 320×320 tiling and rembg segmentation to isolate the airframe. Diagnosed and fixed a data-leakage issue by moving to aircraft-level train/test splits, giving an honest production-split score of mAP@50 0.76 and 0.55 recall. Replaced manual visual inspection with automated defect scoring.
Software Engineer · GSPL
Mar 2025 – Aug 2025Singapore
Produced modular system designs, block diagrams and specs, and wrote API documentation with GovTech and iWOW. Built AWS-backed services (EC2, Lambda) and developed Buddy of Parents (BOP), a React Native mobile app.
Software Lead · Singapore Armed Forces
Nov 2023 – Mar 2025Singapore
Led the full-stack build of Social40, a unit ERP, from scratch in Next.js with SSR. Automated CSV/data workflows in Python and restructured logistics databases to cut redundancy.
Projects
Selected work
AI/ML · Computer Vision
SIA Paint-Defect Detection
May 2026 – Present
A YOLOv8 pipeline detecting paint-peel defects on aircraft fuselages across the Singapore Airlines and Scoot fleets. rembg segmentation isolates the airframe, SAHI 320×320 tiling preserves small defects, and aircraft-level train/test splits keep evaluation honest - mAP@50 0.76 on the production split. Inference scores each aircraft across 15 fixed camera views and publishes annotated images and Excel reports to SharePoint.
Read the case study →AI/ML · Audio
GenreAI
Jun 2024
A neural-network music-genre classifier built with TensorFlow/Keras. librosa extracts MFCCs, chroma, spectral and tempo features from raw audio, and the model classifies tracks across the GTZAN dataset's 10 genres - reaching 99.67% accuracy after extensive hyperparameter tuning.
Read the case study →Full-Stack · Flagship
Gossip with Go
Dec 2025 – Jan 2026
A full social forum - posts, topics, likes, reposts and two-level nested comments - with algorithmic content ranking and multi-entity follows. The Go REST API (25+ endpoints, 3-layer architecture, GORM) is tuned with composite indexes and denormalisation for sub-100ms reads; auth is JWT in httpOnly cookies with rate limiting and bcrypt. Shipped as an installable PWA.
Read the case study →Full-Stack · ERP
Social40
2023 – 2025
A unit ERP for the Singapore Armed Forces combining admin management - personnel records, duty scheduling, attendance, approvals and analytics - with an internal social layer. Built in Next.js 14 (App Router), TypeScript and Sentry monitoring across client, server and edge.
Read the case study →E-Commerce
Mademoiselle Angel
Sep – Oct 2022
A full-stack e-commerce storefront - catalogue, cart and checkout - integrating Stripe and PayNow on a Node/Express backend with Firebase.
Read the case study →Skills
What I work with
AI Ecosystem Tools
- Claude Code
- Cursor
- Codex
- Claude in Chrome
- MCP integrations (Gmail, Drive, GitHub)
- Higgsfield
- ElevenLabs
- Make
- n8n
- Kaggle
- Google Colab
AI Libraries
Languages
Frameworks
Infrastructure & Data
Awards
Recognition
SMU @ Computing Award
2022Top 16 · DSTA × HCI Cyberthon
2022Exemplary Distinction · Canadian Computing Competition (U. Waterloo)
2022Bronze · National Olympiad in Informatics (NUS)
2022