AI / Computer Vision Researcher · IIIT Nagpur
Building physics-informed ML systems and agentic AI platforms.
I'm an AI researcher interested in the intersection of physics, cognition, and machine learning. Most of my work tries to answer one question: can we build models that reason from first principles rather than memorize patterns?
- 🔬 Currently researching physics-anchored deepfake detection (PRISM framework)
- 🧬 Built an agentic AI tutoring platform with epistemic state modeling
- 📷 Interested in sensor physics, noise modeling, and interpretable computer vision
- 🎓 Based at IIIT Nagpur, India
Languages: Python · TypeScript · C++ ML / CV: PyTorch · OpenCV · scikit-learn · NumPy · SciPy Web: Next.js · React · Tailwind · Node.js Tools: Git · Docker · SQLite · Vercel · Claude API
| Project | Description | Stack |
|---|---|---|
| 🛡️ AEGIS | Agentic AI learning platform with hierarchical memory, Ebbinghaus forgetting curve, and Theory of Mind modeling | Next.js · TypeScript · Claude · D3.js |
| 👻 Phantom Lens | Physics-anchored deepfake detection with 24 feature pillars (shadow geometry, sensor noise, boundary analysis) | Python · PyTorch · OpenCV |
| 🔭 PhysDenoiser | Image denoiser trained on physically realistic Poisson-Gaussian noise. No pretrained weights, no APIs — from scratch. | PyTorch · NumPy |
"The best way to predict the future is to build it — one physically-grounded model at a time."