🎓 Integrated M.Sc. Computer Science · Central University of Rajasthan · Batch 2021-2026
🏛️ Research Intern @ INCOIS (Indian National Centre for Ocean Information Services)
🌊 Focused on AI for Oceanography — super-resolution,sea surface temperature forecasting, deep learning for remote sensing
💡 Passionate about applying Super-Resolution and foundation models (Moirai, Chronos, Granite) to real-world scientific problems

ML/DL: Deep Learning, CNNs, RNNs/LSTMs, Transformers, GANs, Time Series Forecasting
Foundation Models: Salesforce Moirai, Amazon Chronos, IBM Granite TSFM
Domains: Computer Vision, Satellite Imagery, Oceanographic AI, Super Resolution

LSTM · N-BEATS · Salesforce Moirai (55M) · Novel 4-Stage Post-Processing Pipeline

Fine-tuned Salesforce Moirai on 16,300 days of OISST data to achieve 0.108°C RMSE — the best Arabian Sea SST forecast reported to date. Introduced a novel 4-Stage Post-Processing Pipeline (Bias Correction → Spatial Correction → Gated Scale → Trend Nudge) that reduced RMSE by 0.013°C.

python pytorch moirai n-beats time-series oceanography incois

ConvLSTM · Amazon Chronos (200M) · IBM Granite TSFM (71K) · Five-Gate Evaluation

Benchmarked zero-shot foundation models against a trained ConvLSTM over a 60×48 Arabian Sea grid. IBM Granite (71K params) achieved 0.1196°C RMSE — outperforming ConvLSTM (0.1417°C) by 15.6% in zero-shot mode. All models validated against 37 Argo float profiles.

python pytorch amazon-chronos ibm-granite convlstm deep-learning sst incois

Deep Learning · Image Upscaling · Remote Sensing

python deep-learning computer-vision super-resolution

"Making the invisible ocean visible — one model at a time."