At IIT Bombay, I work on DhatuVerse geological survey datasets, building end-to-end machine learning pipelines for mineral prospectivity prediction. Through feature engineering and hyperparameter tuning, I raised the prediction F1-score from around 30% to over 70%.
I also built a magnetic inversion model to recover 3D subsurface susceptibility and containerized modeling workflows with Docker, serving them through FastAPI endpoints. For sparse drill-hole grids, I implemented isotropic, anisotropic, and cubic interpolation.
My research includes Halosight, a multimodal deep learning system for detecting Earth-directed Halo CMEs from Aditya-L1 data. The work fused VELC coronagraph images with SWIS-ASPEX particle streams, received Best Paper, and was published in IJSRSET.
I’ve also trained a YOLOv12 crater-detection pipeline on satellite images and built a gesture-controlled drone simulator using OpenCV and MediaPipe. I’m a winner of Smart India Hackathon 2024 and have mentored students through AI/ML workshops at TechNova Club.
