Hi, I'm Ayan, a robotics and AI researcher with an MSc in Robotics & AI from ITMO University (Saint Petersburg). I build autonomous systems that reason about space, combining environmental perception with motion planning for reliable navigation in constrained environments.
- 🎓 MSc, Robotics & AI, ITMO University (with honours); BE, Electrical Engineering, NED University, Karachi
- 🔬 Thesis: training-free 3D perception for autonomous ground vehicles (YOLOv8-seg + LiDAR fusion, evaluated on KITTI)
- 🧭 Navigation: A*, DWA, MPC, PID, UKF state estimation
- 👁️ Vision: YOLO benchmarking, stereo vision, depth-supervised NeRF
- ⚡ Background in power systems and classical control, which I carried into robotics
- 🌍 Based in Karachi · Open to PhD and research collaborations
| Project | What it is | Stack |
|---|---|---|
| 3D Perception for Autonomous Ground Vehicles (MSc thesis) | Training-free pipeline: YOLOv8m-seg → calibrated LiDAR projection → adaptive mask erosion → DBSCAN → PCA 3D oriented boxes. KITTI: Car IoU 50.51%, Pedestrian IoU 38.63%, ~11 FPS, 1,874 MiB peak GPU memory | Python, PyTorch, open3d |
| TriLayer-Nav | Tri-layer navigation framework (A*, DWA, MPC) for differential-drive robots; Gyroscopy and Navigation | Python, MuJoCo |
| Depth-Supervised NeRF | Comparative study of depth-integration strategies for NeRF training quality | PyTorch |
| YOLO benchmark on KITTI | YOLOv8/11/12 compared on mAP@0.5, mAP@0.5:0.95, FPS, FLOPs, class-wise results | PyTorch |
| Multi-Agent Study & Coding Assistant | 8-agent LangChain/LangGraph system with a Reviewer agent for hallucination detection, evaluated on SWE-bench Lite | LangGraph |
| Citation Reliability of Small LLMs | Verified 400+ generated references via Crossref/OpenAlex/arXiv across 45 topics | Python |
- TriLayer-Nav: A Tri-Layer Navigation Framework Integrating A*, DWA, and MPC for Differential-Drive Robots, Gyroscopy and Navigation (2026)
- Improving NeRF Training Quality Using Depth Data: A Comparative Study of Depth-Integration Strategies for 3D Scene Reconstruction (2026)
- A Hybrid ARIMAX–LSTM Framework for Robot-Assisted PM2.5 Forecasting and Adaptive Environmental Monitoring, Atmospheric Research (under review)
perception · planning · autonomy