Tahmidul Bin Ferdous.
Building machines that see, understand, and act.

Real-time air canvas system using hand landmark detection to track fingertip motion and draw directly on screen via webcam gestures.

Facial movement-based mouse control tracking nose coordinates for cursor movement and eye blinks for clicks, enabling hands-free computer interaction.

10M parameter GPT-1-like autoregressive transformer built from scratch for Bengali poetry generation. Trained on 3M-token corpus with character-level tokenization on Kaggle P100 GPU.

Deep Convolutional GAN trained from scratch with custom training loop and adversarial dynamics to generate realistic handwritten Bengali digits.

Real-time hand tracking drawing app with gesture-based controls for drawing, erasing, undo/redo, and color change via webcam.

YOLOv11-based system to detect, count, and track university transit buses in video feeds for automated traffic monitoring.

Deployed computer vision system for real-time packet counting and inventory automation, delivered within a 15-day contract timeline.