Pranab Mohanty

- Lead the development and optimization of foundational LLMs, ensuring alignment with the strategic goals of capital one technology innovation and delivering exceptional performance. - Spearhead the adoption and integration of generative AI technologies to streamline workflows and enhance customer engagement, prioritizing innovative solutions that drive tangible business outcomes - Spearhead the adoption and integration of generative AI technologies to streamline workflows and enhance customer engagement, prioritizing innovative solutions that drive tangible business outcomes. - Deliver state-of-the-art recommendation systems for capital one products and services, utilizing advanced generative AI algorithms to personalize customer experiences and optimize product offerings. - Cultivate and lead high-performing teams focused on the practical AI/ML application and integration of generative AI technologies, mentoring team members to foster professional growth and collaboration. - Partner with product, technology, and platform teams to ensure cohesive integration of AI tools and systems, promoting synergy across lines of business. - Champion research initiatives and publications to address real-world challenges, driving the advancement of state-of-the-art LLMs and AI assistants that meet evolving market demands. - Drive Safety AI innovations (guardrails, AI ethics etc) ensuring the ethical, safe, and trustworthy application of AI technologies across the organization.

• Led the team to optimize and scale Llama2 and multi-modal large language model (MM-LLM) through fine-tuning, instruction-tuning, RAG, DPO, RLHF and scaled to product integration. • Drive development of safety, privacy, and integrity solution for LLM/MML model to make the smart glass compliant with end-user policy. • Drive core technology development for multi-modal chatbot in Ray-ban Meta smart-glasses • Successfully launched 0-1 initiatives and products for a new customer base, navigating ambiguous circumstances including market analysis, launch metrics and post launch improvements. • Led and supported teams in developing virtual assistant features for Meta 1P products (Portal, Quest, etc.), resulting in increased user retention and adoption. • Spearheaded the development of a personalized search and recommendation system for Meta Quest (Oculus) using Online Learning (MAB) and Graph Embeddings.

Leading various incubation and AI initiatives at the center of Excellence in AI at Fidelity.

Managed and Lead a team to build smart and transparent data migration service to Azure Cloud powered by predictive models and recommendation systems and Intelligent Database in Azure Cloud using Machine Learning technology

• Intelligent QnA System: Managed and Lead a Data Science team responsible for building an intelligent Question/Answer system for deep technical queries. The back-end system powered by Microsoft chat bot framework helps Microsoft Azure ML customers find answers to technical issues automatically and reduced support engineer engagement by 30-40%. The intelligent system utilizes cutting edge deep learning technology along with a knowledge base build on technological entities. [2016, Microsoft, USA] • Social Media Analytics: Managed and Lead a Data Science team to build a live Insight framework from Social Media posts, comments, tweets, blogs from many social channels such as Twitter, StackOverFlow, Reddit, HackersNews, MSDN forum and many other forums. The tool is empowered with various NLP components such as sentiment analysis, topic modeling and detection, Author reputation, Post categorization, intent detection etc. [2015-2016, Microsoft, USA] • Anomaly Detection in Time Series: Developed a customized solution for anomaly detection and loan risk assessment for a financial organization as part of Microsoft Customer engagement and empowerment program. [2016, Microsoft, USA] • Ticket Clustering and Classification: Build an end to end solution for our support team to cluster, categorize and correlate troubleshoot tickets which help support engineers to find similar resolutions and reduce to time to mitigate customer issues. [2015, Microsoft, USA] Key Areas: Big Data, Predictive Modeling, Machine Learning, Social Network Data Analysis, Natural Language Processing

• Amazon Go Technology: One of the early member of the research scientist in Amazon Go Technology from its incumbent to private preview. I lead the research on friction free entry of customers to the store, sensor calibration and alignment, sensor fusion, optimal sensor placement and organization, development of in-house multi-sensor technologies, product classification through video/image analysis. [2013-2015, Amazon Inc, USA] • Fast and Accurate interpretation of live video streams from multiple sensors • Action recognition and Behavior Analysis form videos • Sensor Calibration, Optimal Sensor Configuration, and Automated sensor quality control system • Stereo Vision and 3D reconstruction • Face/Fingerprint/Iris Recognition system for Person Identification and Person Re-identification

Design and Develop Biometric Application Algorithm development for Face, Fingerprint and Iris recognition Indexing Biometric Templates for efficient retrieval Quality Analysis on Biometric Templates Compliance Analysis of Biometric Template

Project: Learning from Biometric Distances ● Modeling various face recognition system from the match scores with affine transformation from match scores ● Reconstructing face templates from match scores ● Indexing Biometric Templates ● Exploring privacy and security issues in Biometrics System Project: Analysis of Uni-modal and Multi-modal Biometrics in Challenging Environments

Was associated with RETINA project

Face Recognition, Biometrics, Image Processing, Computer Vision, Patter Recognition

Machine Learning basic Principle Advance Machine Learning and Applications Supervised Learning, Clustering, Ensemble Learning, Decision Trees, Structure Prediction Graphical Models, Support Vector Machine Dimensionality Reduction