Industry Experience

Professional Experience

2024-Present: Sr.Manager, Data Science (AI & NLP), Enterprise Data Science, CapitalOne, Cambridge, MA

Led the IR Team for Agent Assist in developing a state-of-the-art RAG system to support thousands of customer support agents across various lines of business.
Enhanced IR systems by integrating cross-encoder re-rankers, hybrid search methodologies, and dynamic configurations. Key optimization strategies included user preference re-ranking, ranked fusion, entropy-based selection of top preferences, and nucleus thresholding for retrieval.
Directed high-performing teams in designing, implementing, and optimizing deep learning models, employing advanced techniques such as model perturbations to improve robustness, generalization, and adversarial resilience.
Managed end-to-end NLP projects, from conceptualization and data acquisition to model training, evaluation, and deployment, ensuring effective utilization of LLMs.
Developed and fine-tuned deep learning models for question answering and text generation, focusing on scalable and efficient NLP system deployment.
Collaborated with cross-functional teams—including researchers, data scientists, software engineers, and business stakeholders—to translate complex business requirements into actionable technical strategies, ensuring successful project execution and impactful results in AI, NLP, and deep learning.

2022-2024: Manager, Data Science (AI & NLP), Enterprise Data Science, CapitalOne, Cambridge, MA

Lead the IR Team for Agent Assist to built the SOTA RAG system for helping thousands Customer Support Agents across several LoBs.
Lead high-performing teams in the design, implementation, and optimization of deep learning models, leveraging advanced techniques such as model perturbations to enhance robustness, generalization, and adversarial resilience.
Built and managed end-to-end NLP projects, from conceptualization and data acquisition to model training, evaluation, and deployment, ensuring the effective utilization of LLMs.
Built several deep learning models in the area of question answering, text generation, with a focus on implementing and fine-tuning LLMs, deploying scalable and efficient NLP systems.
Collaborated with cross-functional teams, including researchers, data scientists, software engineers, and business stakeholders, to translate complex business requirements into actionable technical strategies, drive successful project execution, and deliver impactful results in the areas of AI, NLP and deep learning.

2021-2022: Value Stream Manager and Technology Lead, AI in PMS applications, Philips Research North America, Cambridge, MA

Value Stream Manager for AI and NLP in Post-Market Surveillance(PMS) across Philips (~$2 Million).
Lead cross-cluster product development for Philips’s PMS and Quality & Regulatory organizations.
Lead a team of 6 researchers and engineers to deliver high-quality AI/NLP solutions to all businesses across Philips.
Defining and delivering state-of-art AI/NLP solutions from design to deployment.
Stakeholder engagement and Roadmap creation across businesses and functions inside Philips.

2017-2021: Senior Scientist, Philips Research North America, Cambridge, MA

Co-Architect heterogeneous swarm-based platform for deploying deep-learning-based models at scale.
Lead a team on Question-Answering systems to help users with complex medical devices and functionalities.
Built Question Answering system focussed on the interpretability of the reasoning behind the answers.
Built Visual Question Answering system that contextualizes the question with respect to image, and retrieves answers containing both images and text.
Developed on Knowledge graph driven clinical diagnosis
Strategic Planning
Stakeholders Engagement
New Proposition and Algorithm Ideation and Development
IP Creation

Written more than twenty Invention Disclosures(IDs) with four IDs as the first inventor.
Lead Knowledge graph-based Clinical Question answering project.
Contributed to two technology transfers to business
Participated in three challenges in TREC’16
Contributed to publications in top NLP and AI conferences.

2014-2015: Post Doctoral Research Associate at Pacific Northwest National Laboratory, Richland, WA

Contributed to the streaming interface for data filtering and aggregation Streaming graph search under the Idaho Bailiff Initiative.
Contributed to the public release of the software
Developed query algorithms and middleware interface for querying large Bayesian networks for cyber security.
Contributed to NOUS project: Knowledge graph construction and maintenance
Contributed to project Chiron under the Defence Threat Reduction Agency(DTRA) initiative
Contributed to large scale machine learning algorithms software release MATEX
Some of the technologies used in the above projects include: Spark streaming framework, Hadoop eco-system, Active MQ, Storm, Genie, Smile Libraries Latent Dirichlet Allocation (LDA), Semantic Role Labeling(SRL), Semantic Parsers, and Stanford NLP Libraries.

2013-2013: Data Analyst Intern (ASTRO program) at Oakridge National Laboratory, Oakridge, TN

Analyzed algorithms for getting an average episode-of-care pattern for patients with chronic diabetes from the medical claims database.
Developed co-reference pattern-based storyline detection algorithm for Agatha Christie novels. Analyzed the interplay between protagonist and antagonist in mystery novels by building a co-reference network.
Some of the technologies used in the above project include: Java, Jung, MongoDB and Neo4J

Designed and developed data warehouse tools for holding the criminal data from all the states of the US.
The process includes writing massive computational programs, stored procedures, and transformation XMLS to run the Data Loading Engine, Data Extraction Engine, and Data Transformation Engine.
Integrated Bugzilla and Sugar CRM on the bug module for transparency of software bugs from the customers to IT development team
Technologies used include Php, MySQL, Zend MVC, SOAP, XML-RPC, C#, VB.net, SQL server 2005.