i am a 3rd year b.tech student studying artificial intelligence and machine learning at nmims.
currently i am working on building applied machine learning systems involving spatio-temporal models, multimodal retrieval, and real-world data pipelines.
previously i have worked on problems ranging from air pollution forecasting to climate data systems and infrastructure-related ml problems.
1. pm2.5 forecasting (physics-informed ml): built a convlstm-based spatiotemporal model to predict pollution levels across india with a custom wind advection module. won 1st place at anrf aisehack (iit delhi + ibm research).
2. offline multimodal rag system: built a fully offline system for querying text, images, and audio using faiss, clip, and whisper without relying on cloud apis.
1. floatchat: natural language interface for querying oceanographic climate datasets, reducing manual analysis effort significantly.
2. underground pipe leak detection: ml pipeline using acoustic and sensor data for detecting leaks in pipelines, developed for a smart city initiative.
1. speaker identification (voxceleb): cnn-based model trained on mel spectrograms with augmentation, achieving 86.6% validation accuracy.
- 1st place — anrf aisehack (pollution forecasting)
- 2nd place — yugantar 2025
- round 5 — hackrx 6.0
- finalist — innovent sobus
python, c++
pytorch, tensorflow, scikit-learn
rag systems, bert, langchain
numpy, pandas
flask, fastapi
[+] last updated: april 2026