01 / Thesis
Weather prediction
SFNO-based model, fine-tuned and benchmarked for regional forecasting. Now extending with mechanistic interpretability.
NLP / DLWP / Mech Interp
Vishnu
Kadam
BS-MS graduate, IISER Pune. Deep learning from theory to application — now going deep on mechanistic interpretability.
Now
Investigating mechanistic interpretability techniques on AI weather models — uncovering what these models actually learn internally.
Stack
Publication / 2026
CurvePotGCN — predicting protein–protein interactions using surface curvature and electrostatic potential as node features
read paper →02 / Research
Protein interaction
End-to-end GNN pipeline — dataset creation, architecture, training, benchmarking. My first publication.
03 / Project
Diffusion models
Annealed Langevin dynamics, implemented from scratch with extensive visualisations.