EleutherAI is a leading non-profit machine learning research group and hack lab focused on developing and understanding large-scale ML models, their applications, and ensuring their safety. As part of the team, my work involves advancing research in LLM (Large Language Model) data privacy and exploring novel approaches to reasoning in machine learning systems. I am deeply involved in studying the ethical implications and privacy concerns surrounding LLMs, with a focus on creating safer, more robust AI systems.Key Responsibilities:Conducting research on data privacy in large language models, ensuring models protect sensitive information while maintaining high performance.Investigating approaches to reasoning within LLMs to improve model interpretability and decision-making capabilities.Collaborating with other researchers to publish findings and contribute to open-source AI safety initiatives.Exploring the ethical and societal impacts of large-scale AI models, particularly in high-stakes applications.

At Kaiser Permanente, I focus on Conversational AI and Neural Information Retrieval to enhance healthcare solutions.Conversational AI: Fine-tuned state-of-the-art sequence-to-sequence models such as T5, BART, and LLaMA to develop a context-sensitive medical chatbot. This solution is designed to understand and respond accurately to patient queries using utterances in a medical setting.Neural Information Retrieval: Implemented a sophisticated re-ranking architecture utilizing RAG, LlamaIndex, and LLMs to improve the accuracy and relevance of medical information retrieval in real-time.This work drives better patient experiences through advanced AI and NLP technologies in the healthcare sector.

Team: AI Systems Hardware/Software Co-Design:Role: Benchmarking Automation for Advertisement, Ranking, and Personalization using Sparse Neural Network Models. Focused on Performance Evaluation and Hyperparameter Fine-tuning.Key Responsibilities:Developing and automating benchmarking processes for AI models in advertisement, ranking, and personalization systems.Evaluating performance metrics and fine-tuning hyperparameters for Sparse Neural Networks to optimize results in large-scale systems.Collaborating on AI systems hardware/software co-design for efficient integration and enhanced model performance.

As a Principal NLP Engineer at CVSHealth, I successfully completed several key projects focused on enhancing search functionality and relevance in intelligent systems:Intelligent Search and Question Answering System: Led the development of a sophisticated question-answering system, enabling efficient and accurate information retrieval.Image to Prompts – Stable Diffusion: Worked on generating accurate prompts from images using Stable Diffusion, contributing to enhanced visual understanding.Personalized Search Query Completion: Implemented advanced techniques for personalizing search suggestions based on user behavior, improving search accuracy and user experience.Search Query Classification: Developed query classification and rewriting algorithms to enhance relevance and deliver more meaningful search results.

As a Mentor and Code Reviewer at Udacity's School of AI, I played a crucial role in supporting student success through detailed project reviews and personalized mentorship:Project Reviewer: Reviewed student project submissions, providing actionable and constructive feedback to help students improve their work and understand key concepts.Student Mentor: Guided students through technical course-related questions and challenges, offering support and motivation to help them complete their courses on schedule.Beta Testing: Participated in beta testing of new Udacity products, providing feedback to improve course content and user experience.During my time at Udacity, I specialized in several advanced AI and machine learning techniques:DDPG - Actor-Critic PolicyValue-Based MethodsDeep Q-NetworksOpenAI GymUnity Multi-Agent Systems

Entity Linking, ML Flow pipeline, Feature Store

As a Machine Learning Architect on the Search Team at ServiceNow, I focused on developing cutting-edge solutions in the fields of Semantic Search, Knowledge Graphs, Natural Language Processing (NLP), and Natural Language Generation (NLG). My work involves leveraging advanced Meta Deep Learning Models to drive innovation in search technologies, enabling more accurate and context-aware information retrieval.Key Responsibilities:Developing and optimizing Semantic Search systems to enhance the accuracy and relevance of search results.Building and integrating Knowledge Graphs to improve structured data representation and relationships within search queries.Working with advanced NLP and NLG techniques to enhance natural language understanding and generation capabilities.Utilizing Meta Deep Learning Models to solve complex machine learning challenges in search and AI-driven automation.

As a Staff Machine Learning Engineer at WalmartLabs, I contributed to the development of advanced machine learning models and systems to improve key processes, particularly in attribute extraction, classification, and ranking.Attribute Extraction Models: Developed models for extracting key product attributes, and automation to releasing to productionDeep Neural Networks (CNN & RNN): Built and optimized convolutional neural networks (CNNs) and recurrent neural networks (RNNs) for various predictive tasks.Transfer Learning & Q-Learning: Applied transfer learning to accelerate training across tasks and utilized Q-Learning for reinforcement learning applications.Learning Vector Quantization (LVQ) & Ensemble Methods: Leveraged LVQ and ensemble techniques to improve the accuracy and robustness of classification and ranking models.Word Embeddings & Neural Machine Translation: Worked on word embedding techniques and neural machine translation (NMT) systems to enhance language understanding in various applications.Ranking Algorithms: Developed and fine-tuned ranking algorithms to improve search relevance and recommendation systems.

Internet Of Things project: Retail IOT, Real Time Event Processing ( real time data streaming ) with Kafka and Apache Storm, Cassandra, Java, python

BigData Challenges: Java, Hadoop, Spark

Acquired by HomeAdviserDuring my time at AngiesList, I played a key role in building and optimizing search and recommendation systems, contributing to the company's success before it was acquired by HomeAdvisor. My focus was on leveraging machine learning and search technologies to enhance user experience and business outcomes.Geospatial Search & Recommender Systems: Developed and optimized a geospatial search and recommendation system using Lucene and Solr, providing highly relevant location-based search results.Sentiment Analysis: Implemented Sentiment Analysis on review texts using Deep Neural Networks, enabling better customer understanding and insight extraction.Query Understanding and Rewriting: Designed and deployed advanced query understanding and rewriting mechanisms, improving search relevance and user intent matching.Named Entity Recognition (NER): Applied Named Entity Recognition (NER) techniques to identify key entities from user queries, enhancing query processing and relevance.Query Intent Prediction: Built a Query Intent Prediction model using LSTM to anticipate user needs and deliver more precise search results, improving overall user satisfaction.

At WalmartLab, I contributed to several high-impact projects focused on Web Architecture, Scalability, and Real-Time Analytics, playing a key role in optimizing system performance and delivering data-driven insights for business growth.Web Architecture and Scalability: Designed and implemented scalable web architectures to support high traffic and ensure seamless user experiences. Focused on optimizing performance for large-scale systems.Distributed Caching: Implemented distributed caching strategies to enhance system efficiency and reduce latency, ensuring faster data access and improving response times for users.Real-Time Analytics: Led projects focused on real-time analytics, enabling the business to capture and analyze large-scale data streams instantaneously, driving actionable insights and decision-making.

As a Senior Software Engineer in the Search and Recommendation team, I developed and optimized search functionalities, improving user experience and business outcomes through advanced search algorithms and personalized recommendations.Implemented Automatic IR Evaluation System: Leveraged the MS-MARCO document ranking dataset to enhance search evaluation, real-time indexing, and search ranking algorithms for more accurate and efficient information retrieval.Search Features: Designed and developed key search components, including an Auto-Suggest feature and Sponsored Search functionality to increase engagement and revenue.Web Crawler Implementation: Built and deployed a secure web crawler using NUTCH for indexing static and dynamic web pages, enabling faceted search and browsing across various data dimensions.Search Relevancy Tuning: Developed new ranking algorithms for search relevancy tuning and query understanding with query rewriting techniques to improve search results.Customer Segmentation: Successfully implemented K-Means Clustering for customer segmentation to tailor marketing strategies and recommendations.Deep Personalization: Built a deep personalization ranking mechanism using user search and click history to provide more relevant search results based on user behavior.

Product Development - LeadN-Tier architecture, VC++, Windows, Oracle, Object Modelling and Code Generators

Developed Material Resource Planing software system on Unix/Oracle/C platform.

C-Kit - C Library for GUI on Windows platform