Muhamad
Amirul Aiman

An automated industrial defect-detection system pairing a custom-trained YOLOv8 model with a FastAPI backend and physical conveyor assembly — delivering real-time pass/fail classification at machine speed.

• Trained and deployed a custom YOLOv8 model for high-confidence defect detection on Hicom Diecasting material components
• Engineered a hardware-software handshake achieving sub-200ms detection-to-mechanical response with zero missed detections under load
• Built an automated Supabase data pipeline eliminating manual record-keeping with 100% logging accuracy
• Visualized live inspection metrics via a Streamlit dashboard for real-time operator oversight

Designed and built the serverless backend for a generative AI solution during the VIBE 2026 AWS Ideathon — orchestrating large-model inference through Amazon Bedrock via a Lambda + API Gateway execution path.

• Architected Amazon Bedrock (Nova Pro) orchestration layer for structured generative AI output across complex multi-step workflows
• Built serverless execution paths using AWS Lambda, handling inference triggers without persistent server overhead
• Advanced prompt engineering and AI tokenization optimization to reduce output latency within AWS cost constraints
• Integrated auto-evaluation platforms generating Vite + React frontends via AI prompting in a competitive team setting

A comprehensive pharmaceutical management platform digitizing inventory control, transaction processing, and real-time stock tracking for a simulated drugstore environment.

• PHP, HTML/CSS, and MySQL stack — managing 20+ medicine SKUs with real-time stock updates, reducing manual audit time by 40%
• Secure checkout and billing module processing transactions with 100% financial data accuracy
• Responsive search and filter interface improving pharmacist product retrieval speed by 25%

A desktop-based library management tool demonstrating rigorous OOP design — handling 150+ book records, member profiles, and lending logic with zero-corruption data persistence.

• Applied Encapsulation and Inheritance in Java to model complex lending rules and book/member hierarchies
• Automated fine-calculation algorithm improving policy enforcement consistency by 20%
• Custom error-handling suite reduced system crashes by 15% during final testing phase