My research focuses on three-dimensional mirror symmetry — classifying dual pairs in supersymmetric quantum field theories with minimal supersymmetry, an open problem for three decades. My doctoral work developed a systematic framework for constructing these dual pairs, with findings published in Physical Review D and the Journal of High Energy Physics.
📄 Find my publications here
Alongside the PhD, I have been building end-to-end ML engineering skills: agentic workflows, retrieval systems, geometric deep learning, time series forecasting, and domain-adaptive LLM training. The repositories below are the output of that work.
Languages
Data Handling
ML & Deep Learning
LLMs & Agentic AI
MLOps, Deployment & Tooling
Spectral Response Fidelity · JAX FNO study of whether neural operators learn physically meaningful response functions; spectral fidelity improves with supervision of Jacobians.
Domain-Adaptive LLM Trilogy · RAG → GraphRAG → QLoRA pipeline for theoretical-physics knowledge, culminating in a domain-adapted Qwen2.5-3B.
Low-Resource NMT · MarianMT EN→ML fine-tuning, improving BLEU 15.48 → 32.61 on technical-domain data.
Agentic AutoML · LangGraph + MCP system for autonomous supervised ML with Docker, MLflow, and LangSmith.
NL-to-SQL Agent · Retrieval + execution-guided self-correction; 54.5% BIRD Mini-Dev accuracy with 55% fewer tokens. 🤗 demo
Multi-Agent Symbolic Reasoning · Solver → verifier → critic architecture with SymPy verification via MCP. 🤗 demo
MolPredict · GATv2 molecular property prediction deployed as a FastAPI + MCP inference service. API
Orbit Wars · GATv2 + PPO with action masking, differentiable graph construction and curriculum learning.
Time Series Forecasting · LSTM, Transformer, AutoTFT and XGBoost across BTC/USDT and Indian electricity-demand forecasting, including climate stress tests.
