- 🎓 B.Tech in Computer Science & Engineering, Indian Institute of Technology Jodhpur (CGPA: 8.39)
- 🔬 Undergraduate Researcher, CFD–Machine Learning Group, Dept. of Aerospace Engineering, IIT Jodhpur — building physics-aware neural operators for aerodynamic flow prediction and trustworthy RL frameworks for risk-averse aerospace design optimization
- 🌐 Research Intern, Numerical Analysis Group, TU Delft — domain decomposition & Krylov solvers for multiphysics PDEs
- ✍️ First-author paper submitted at journal (under review) — preprint available on arXiv
- 🧠 Interests: Scientific Machine Learning · AI4Science · Multiphysics & Multiscale Modelling · Cyber-Physical Systems & World Models · Embodied AI and Robot Learning · Reinforcement Learning · Autonomous Systems and Robotics · AI for Defence
- ⚙️ Also love exploring and building projects in AI, embedded software and system design on the side and expanding my knowledge (RAG pipelines and agentic AI, low-level programming for microcontrollers)
Some of my research project repos are still private; they will be made public as soon as I publish.
AI/ML and Scientific Software
Languages
Full-Stack & Infra
Robotics & Embodied AI
| Focus | Description | Links |
|---|---|---|
| Domain-Decomposed Reduced/Neural Operators (DD-RNO) (IIT Jodhpur) | Physics-aware neural operator for 2D airfoil flows, conditioned via SDF geometric encoding — first-author, under journal review | Preprint · Repo |
| RL-assisted PDE-Constrained Shape Optimization (IIT Jodhpur) | RL framework using neural surrogates as black-box evaluators for aerodynamic shape optimization | — |
| Multiphysics Domain Decomposition (TU Delft) | Scalable algorithms for multiphysics and multiscale simulations | — |
| Project | Description | Repo |
|---|---|---|
| ML-Driven Aerodynamic Design (STOL Aircraft) | Surrogate-assisted optimization framework (OpenFOAM + XFOIL/XROTOR + Differential Evolution) — 20–30% drag reduction, Inter-IIT Tech Meet 14.0 | link |
| NeuroSINDy | Extended SINDy framework (TVDiff + Ensemble SINDy + BIC model selection) — 19.4× lower derivative estimation error vs. finite differences | link |
| Wormbrained | Biophysical C. elegans-inspired agent: spring-mass Stokeslet fluid model + 48-neuron LIF spiking controller for chemotactic navigation | link |
| CortexFlow | AI-powered cognitive monitoring platform — Chrome extension + Django backend + React dashboard predicting attention breakdowns from behavioural telemetry | link |
| SignalSight | CNN-based wireless modulation classification on RadioML with Grad-CAM explainability, deployed via Flask API | link |
| ASCEND | GPS-denied UAV navigation & precision landing using VIO/VSLAM + IBVS, built for ISRO's IRoC-U 2026 | no public repo |