AI Systems • Geometric ML • Physical-World Intelligence
Building machine learning systems that survive contact with reality.
Email • LinkedIn • Google Scholar
Final-year Computer Science undergraduate working on geometry-aware AI systems,
multi-modal sensing (WiFi, radar, UAV), and safety-critical real-world ML.
Currently an Undergraduate Research Assistant at IIT Delhi,
building drone-based 3D terrain reconstruction and AI-driven slope risk prediction systems.
I care about:
- Models that generalize beyond benchmarks
- Systems that operate under physical constraints
- Efficient architectures for edge deployment
- Real-world evaluation, not just leaderboard performance
🛰 Geometry-Aware 3D Reconstruction
Photogrammetry • Mesh optimization • Curvature estimation • GIS-integrated modeling
📶 WiFi-Based 3D Human Pose (LightPoseNet)
- 40mm MPJPE
- 93.31% PCK@0.10
- 300× parameter reduction
- Real-time edge deployment
🛩 Radar-Based UAV Classification
- EfficientNetB3 fine-tuned
- 97.10% classification accuracy
- Micro-Doppler drone vs bird separation
🧠 Robust & Safety-Critical ML Systems
- AAAI 2026 — EGSAI Workshop
- IEEE MIT URTC 2025 — Paper Accepted
- IEEE ICEdge 2025 — IISc Bangalore
- University of Cambridge — AI & Cognitive Science
- EAIC 2025 — Radar-Based UAV Classification
- ACM India COMPUTE Fellow (2025)
- Oxford Machine Learning Summer School
- IEEE–EURASIP S3P (Italy)
- IISc EE Summer School
MIT Hacking Medicine 2026
TreeHacks (Stanford)
MIT iQuHACK
Multiple national hackathon finalist finishes
“Not building demos. Building systems.”