CV

eBay

Applied Researcher 1 (SE3) · Knowledge Extraction for Search

• Drove adoption of small multimodal model (SLM)-based generative information extraction at scale (100M+ listings/month), designing and building pipelines that improve on legacy NER and dictionary-based methods.
• Built multimodal agents and workflow pipelines that generate synthetic training and evaluation data using open-source large language and multimodal models (LLMs, LMMs) to accelerate data and model iteration.
• Developed deep learning models for classification, confidence scoring, and bounding-box detection to augment the LLM/SLM information extraction pipeline, improving extraction accuracy and field coverage.

Texas Instruments

ASIC Digital Design Engineer

• Physical design for four taped-out power management ICs: timing closure (STA), power, and EM/IR reliability signoff.
• RTL design of custom floating-point multipliers and a system ALU achieving 40% area reduction versus comparable Cadence IP; plus RTL for PMBus protocol and GPIO control IP.
• Automated physical design flows in Python/TCL, cutting tape-out time by 2×.

University of California San Diego

MS, Computer Science & Engineering

AI/ML specialization. Probabilistic reasoning, reinforcement learning, deep generative models, and recommender systems. Research with Julian McAuley’s group; teaching assistant for recommender systems and data mining.

National Institute of Technology Karnataka

B.Tech, Electrical & Electronics Engineering

Power-quality classification with deep learning, Kaggle competitions, and the Amateur Astronomy Club.

First place · EAI Challenge

Embodied Agent Interface Challenge at the NeurIPS FMEA Workshop. Team AxisTilted2.

Honorable Mention · UIST

ZINify, Student Innovation Contest.

First of 591 teams · eBay University ML Challenge

2022 competition; results announced January 2023.

Best Paper · IEEE DISCOVER

Power quality event classification using LSTM.

Kaggle Competitions Expert

Silver: 98th of 2,380 teams (top 5%). Bronze: 312th of 4,539 teams (top 7%).