I'm a computer engineering student who builds practical vision tools — from multi-camera capture systems to deep learning pipelines for action and gesture recognition. I care about things that work on real hardware, not just in notebooks.

Currently focused on multi-device camera orchestration and video-based gesture recognition.

DroidGrid — Multi-phone DroidCam controller
Preview, record, and snapshot from up to 10 Android phones simultaneously. Self-healing streams, non-blocking I/O, inline session editor. Part of Vision-Orchestration.

Foot Recognition — Industrial foot gesture recognition
Vision pipeline for 8 foot gesture classes in production line environments. Multi-view custom dataset, EfficientNetV2-S + R(2+1)D comparison, structured for publication.

Folio Finder AI — Fall detection from video
R(2+1)D-18 trained on a custom dataset (~7,000 clips). 98.71% F1 on a held-out test set. Full training pipeline with AMP, class weighting, and early stopping.

languages  = ["Python", "Bash", "C/C++ (Arduino)"]
ml         = ["PyTorch", "OpenCV", "EfficientNet", "R(2+1)D", "ConvNeXt"]
domains    = ["Video classification", "Gesture recognition", "Fall detection",
              "Multi-camera systems", "Dataset pipeline"]
tools      = ["Git", "CUDA", "ffmpeg", "SQLite", "tkinter"]
hardware   = ["RTX 3070", "Ryzen 7 5800H", "32GB RAM"]

• Video representation learning and temporal modeling
• Structured evaluation and ablation studies for CV papers
• Edge deployment for vision models (ONNX, TensorRT)