I’m a machine learning engineer/researcher with interests spanning vision-language models, graph representation learning, and optimization for modern deep learning.

CV Projects Publications

What I’m working on

  • AI Health Analytics @ Symviq (Aug 2024–present): vision-language modeling for chronic kidney and heart disease classification from retinal fundus images; building foundation models and CLIP-style encoders for medical imaging.
  • Open-source: building a Keras-first Graph Neural Networks library (GNN-Keras3) with modules like GCN, GAT, and differentiable pooling.

Research interests

  • Vision-language and contrastive learning (CLIP-style training, multimodal retrieval/grounding)
  • Graph representation learning (GNN architectures, pooling, scalable training)
  • Optimization + theory for ML (asynchronous/distributed SGD, PL functions)

Highlights

Publication

  • Convergence of Asynchronous Stochastic Gradient Descent for Polyak-Łojasiewicz Functions (submitted to IEEE Transactions on Automatic Control). PDF · Code

Selected projects

  • GNN-Keras3: Graph neural network training/inference library in TensorFlow/Keras. Repo · Docs
  • Text-to-Image Synthesis Optimization: optimized CBAM + CLIP conditioning under hardware constraints (UCLA MS project).
  • Mini-GPT (TinyStories): small GPT implementation for story generation with parameter-efficient weight tying.
  • VOC signature clustering: ML pipeline for E-Nose VOC signatures (B.E. thesis).

Education

  • MS, Electrical & Computer Engineering (UCLA) — 2023–2024
  • M.Tech, Electrical Engineering (IISc) — 2021–2023
  • B.E., Instrumentation & Electronics Engineering (Jadavpur University) — 2017–2021

If you’d like the detailed version (with milestones, metrics, and links), see Research and Projects.