Building LLM-powered agents at Walmart ASR (formerly Alert Innovation).

Member of the DARPA SemaFor program at BU where we focus on solving real-world problems in generative models and their application in media.• Won 1st place in the competitions for machine generated text detection in news and social media.• Was a key contributor to solving 13 competitions in the SemaFor Project that span detection, attribution, and characterization of NLP tasks. Our team regularly outranked all other performers.• Introduced a new dataset of ~1M real and synthetic news articles and social media posts. The dataset covers various fine-tuned generators and decoding strategies.• Integrated, maintained, and supported API compatibilities for ~20 Docker containers.

Projects highlights:• Developed the server components of the Reliance on Science in Patenting project that retrieves, validates, and analyzes patents data of ~500GB of PDFs and texts.• Benchmarked traditional computer vision and deep learning methods for document layoutanalysis of documents of non-English languages.• Published a joint paper on using machine learning methods to predict Aphasia recovery with the Aphasia Research Laboratory.• Wrote a term paper introducing a Transformer-based contrastive predictive coding architecture for downstream audio tasks.

Worked with the data foundation team on data integration, validation, and query optimization for clients in the Banking sector.