I build end-to-end robotics and intelligent-system prototypes, from sensor firmware and serial protocols to desktop interfaces, computer-vision pipelines, machine-learning experiments, and control-system analysis.
My current focus is turning university prototypes into reproducible, well-documented engineering projects with clear setup instructions, honest limitations, and evidence-backed results.
STM32F411RE firmware and a .NET UWP dashboard with JSON serial telemetry, SQLite history, configurable alarms, and physical LED/servo control.
| Real-time PyQt dashboard combining YOLO face detection, FaceNet identity embeddings, MediaPipe gestures, attention cues, capture, and recording.
|
Regression study of 226,346 cleaned resale transactions. The best recorded Random Forest result achieved test R² 0.9522 and RMSE S$41,311.
| Custom chassis design, fabrication documentation, and Arduino firmware for line following with ultrasonic stop-and-retreat behavior.
|
Two simulation scenarios covering priority-based behavioral autonomy, internal needs, remembered targets, and OpenCV-based object search.
| Pendulum parameter identification, damping analysis, transfer-function and state-space models, stability margins, and lead compensation.
|
- STM32 Robot Navigation Experiments - three independent Mbed applications for Bluetooth control, IR sensing, and ultrasonic obstacle response.
- Chess AI GUI - portable Python/Tkinter chess application with corrected alpha-beta search and optional Stockfish integration.
- Raspberry Pi Sensor Hub - stationary surveillance testbed with reconciled GPIO mappings, modular sensor demos, and hardware-safe initialization.
- Electronics Prototyping Labs - curated Arduino, digital-logic, Multisim/Ultiboard, H-bridge, and timer experiments.
- Arduino Surveillance Patrol Robot - calibrated square patrol with PIR alerts and ultrasonic obstacle response.
- Speaker Identification with MFCC, k-NN, and GMM - MATLAB audio-classification pipeline with frame- and file-level voting.
- Connect software decisions to the real hardware, data, or simulation assumptions behind them.
- Document reproducible setup steps, interfaces, safety constraints, and known limitations.
- Keep private datasets, biometric material, licensed coursework assets, and generated output out of public repositories.
Explore the pinned repositories below for implementation details, results, and build instructions.