Adarsh Vishwakarma

I am a theoretical physics PhD researcher with practical expertise in Machine Learning. My research focuses on high-energy physics and using AI for science. Beyond research, I design machine learning architecture and orchestration pipelines, currently focusing on predictive models for flight delay risk.

Research & Engineering Focus

Physics & AI for Science

My background includes research on soft photon theorems in quantum field theory and co-authoring a publication on high-energy physics (arXiv: 2601.00336). I am currently applying machine learning to theoretical physics, specifically exploring how transformer architectures can be utilized to simplify complex mathematical expressions that appear in Feynman integrals.

Applied Machine Learning

I build practical AI tools, including developing a local RAG engine utilizing Qdrant and FastAPI. My ongoing work involves training and fine-tuning large language models for scientific problem-solving. Additionally, I maintain a strong research interest in mechanistic interpretability to better understand model internals.

Data Architecture & ML Pipelines

Architecting ETL pipeline that includes orchestrating Medallion data pipelines on Google Cloud Storage to process millions of records and integrate data from an API source for predictictive models.

Technical Ecosystem

  • Languages: Python, C++, SQL, Bash
  • Machine Learning & AI: PyTorch, Qdrant, PyDantic
  • Data & Backend Engineering: FastAPI, Airflow, Docker, GCP, Medallion Architecture
  • Applied Mathematics: Numerical Analysis, Mathematica, Sympy