Graduate student at UCLA

AI + Science

I’m interested in building AI and in using it to solve scientific problems.

GitHub

Google Scholar

About

I’m pursuing a Master of Quantum Science and Technology at UCLA. My current work focuses on deep learning and generative modeling for scientific inverse problems, especially ultrafast pulse retrieval using learned priors and flow matching. More broadly, I’m interested in building AI systems and applying them to scientific problems. Previously, I worked on legal AI at Regology, analytics infrastructure at PrashantAdvait Foundation, and distributed machine learning at KAUST.

2025 - Present

University of California, Los Angeles

Graduate Research Assistant · Quantum Light Matter Cooperative

Master of Quantum Science and Technology

Developing deep learning and generative modeling methods for scientific inverse problems, with current work on ultrafast pulse retrieval using learned priors and flow matching.
Implementing Randomized and Deterministic Benchmarking protocols for single and multi-qubit gates on NMR and IBM quantum processors via Qiskit, characterizing gate fidelities and error rates.

2024 - 2025

PrashantAdvait Foundation

Senior Software Engineer

Architected a Couchbase to ClickHouse ELT pipeline via materialized views, enabling sub-100 ms dashboard latency and sub-1 minute refresh SLAs for outreach operations handling 100K+ calls per day.

2021 - 2024

Regology, Inc.

Senior Data Scientist

Technical lead for Reggi, a RAG-based legal AI assistant, driving core design across ingestion, hybrid retrieval, and answer generation while consolidating fragmented regulatory research workflows for enterprise law librarians.
Architected a two-level hierarchical search system in Elasticsearch, combining BM25 lexical retrieval and dense semantic embeddings for 1M+ legal documents across corpora, improving Recall@10 by 5.7x.
Developed a custom keyphrase extraction algorithm for hierarchical legal documents, improving downstream Recall@10 by 1.8x.

2019 - 2021

King Abdullah University of Science and Technology

Graduate Research Assistant · SANDS Lab

M.S. in Computer Science

Proposed an error minimization framework for gradient sparsification, proving hard-threshold sparsification with error feedback is communication-optimal and validating it on multi-GPU clusters.
Contributed to OmniReduce, efficient sparse collective communication primitives that exploit gradient sparsity to maximize effective bandwidth use, accelerating distributed training by up to 8.2x.

INRIA · 2020

Summer Research Intern · STARS Team

Designed a novel ordering-loss framework for weakly supervised action segmentation, using ordered action pairs to model temporal structure without pseudo-label generation or costly Viterbi-based training.

2014 - 2018

Indian Institute of Technology, Kanpur

B.S. in Mathematics and Scientific Computing

Minors in Algorithms and Machine Learning

KAUST · 2017

Summer Research Intern · Optimization and Machine Learning Lab

Analyzed asynchronous parallel SGD for consistent linear systems, proving global linear convergence and showing it can outperform synchronous SGD in ill-conditioned regimes.