Computer Science (AI) student at the University of Nottingham Malaysia (Class of 2027) who ships systems that real people use — from production ERP platforms to medical imaging and site-safety computer vision.
- 🤖 Final Year Project (in progress): an agentic AI order-fulfillment system — LangGraph orchestration + local LLM turning WhatsApp/email orders into validated ERP transactions, with safety-first guardrails (the LLM never decides stock, computes totals, or commits orders)
- 💼 Part-Time Software Engineer @ Summit Company (M) Sdn Bhd — maintaining the 4 production ERP modules I built as a Software Engineer Intern (they kept me on after the internship)
- 📄 First & corresponding author of an ML evaluation manuscript under review at Frontiers in Big Data — auditing benchmark contamination in an oral-health screening pipeline
- 🥽 Built an AR site-safety system as a summer research intern — two internships, one summer
- ☁️ Currently studying for AWS Certified Cloud Practitioner
AI / ML: Python · PyTorch · OpenCV · scikit-learn · YOLOv8 · LangGraph · Ollama · Hugging Face Transformers · RAG Full-Stack: PHP · Laravel · Vue.js · Next.js · Nuxt · TypeScript · JavaScript · Tailwind CSS · MySQL Backend & DevOps: FastAPI · Docker · Nginx · ONNX · Git · GitHub · Linux
| 🧠 AI / ML | |
| 🌐 Full-Stack | |
| ⚙️ Backend & DevOps |
|
AR site-safety intelligence: PPE detection at 98.2% mAP@50 (42k images), real-time AR overlay at ~57 FPS, and a custom identity layer lifting worker re-ID to 100% recall with 0% false merges even in identical PPE — up from an 8% baseline. Deployed to edge via ONNX + an installable phone app. Summer Research Internship · UNM · 2026 |
"Smart Triage" multi-model screening — ResNet-18 router, YOLOv8+SAHI hygiene specialist, multi-task DenseNet-169 pathology expert (96.33% accuracy, 100% recall). Full-stack deployment with RAG medical chatbot and 4-language UI. Basis of my first-author manuscript under review at Frontiers in Big Data. Research & Full-Stack · 2025–2026 |
|
The first large-scale parallel dataset for Bahasa Melayu programming instructions — 800k+ instruction pairs across 12 languages, quality-tiered with BERTScore & COMET and validated against HumanEval-XL. Fine-tuned CodeT5 for Malay-language code generation. Lead Researcher · Summer Internship · 2025 |
Production-shaped publishing platform: Laravel 13 JSON API + Nuxt 4 SSR as decoupled apps. Markdown authoring with safe server-side rendering, draft/scheduled publishing, Sanctum cookie auth, HaveIBeenPwned password screening, full PHPUnit feature suite. Full-Stack · 2026 |