You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am a Systems Programmer, AI Engineer, and Full-Stack Developer passionate about building highly polished, performance-oriented Windows desktop utilities, deep-learning models, and responsive web systems. I specialize in offline-first architectures, local AI deployment, and constraint-based algorithms.
Deep-learning-powered crowd estimation system countering Western data bias.
Features: High-accuracy density estimation + interactive canvas-based crowd flow simulator.
Tech Stack: CSRNet (PyTorch), FastAPI, Vite + Canvas, Google Colab (T4 training).
Highlights: Outperforms original CSRNet paper benchmarks on ShanghaiTech (Part A: 63.31 vs 68.2 MAE | Part B: 8.37 vs 10.6 MAE) by fine-tuning on a custom-curated Indian crowd dataset (Part C).
Features: Set images, GIFs, videos, and live HTML pages as active wallpapers; 3-layer widget overlays, drag-and-drop widget arrangement, click-through toggles.