I am a Machine Learning Engineer on the Adobe GenStudio team in San Jose, where I work on generative AI systems for marketing content. My current work includes models and agents for creative generation, personalization, factuality, and brand adherence.
Previously, I was a research intern in Adobe Research's Video AI Lab and an NLP Research Engineer at ByteDance AI Lab in Singapore. I received an M.S. in Intelligent Information Systems from Carnegie Mellon University, where I worked with Shinji Watanabe and Graham Neubig, and a B.Tech. from IIT Madras.
I am interested in generative models, open-ended learning, and evaluation. Much of my recent work studies how models can create, judge, and improve work when quality is subjective, contextual, or difficult to verify. I also work on robustness in multimodal models and multilingual speech systems.
AAAI Workshop on Creative AI for Live Interactive Performances, 2026 (oral)
A curiosity-based objective helps language-model judges adapt to individual evaluators' subjective assessments of creative writing.
ICLR Workshop on Scaling Self-Improving Foundation Models without Human Supervision, 2025
Investigates self-correction as a route to improve the robustness of multimodal language models under distribution shift.
Interspeech, 2024 (oral)
A benchmark for multilingual speech models across downstream architectures, fine-tuning choices, efficient adaptation methods, languages, and datasets.
ICASSP, 2023
Introduces Visual-ASR-EC and methods that use image information to resolve speech-recognition errors that text alone leaves ambiguous.
NeurIPS Demonstrations Track, 2021
A neuro-symbolic framework for automatically evaluating explanations of graph neural network predictions.
I have reviewed for NeurIPS, ICLR, and workshops at NeurIPS, ICLR, EMNLP, and ACL.
