Data isn’t units of information. Data is a story about human behavior – about real people’s wants, needs, goals and fears. Never let the numbers, platforms, charts and methodologies cloud your vision. Our real job with data is to better understand these very human stories, so we can better serve these people. Every goal your business has is directly tied to your success in understanding and serving people.” — Daniel Burstein

About Me

Combining a passion for innovation with technical expertise, I thrive at the intersection of data science, machine learning, and generative AI. With a Master’s in Data Science from Northeastern University (GPA: 3.9/4.0) and hands-on experience leading impactful AI initiatives at Fidelity Investments and Ernst & Young, I specialize in building systems that turn complex data into actionable insights.

I stay at the forefront of advancements in Generative AI, ensuring I am well-equipped to navigate and contribute to the rapidly evolving landscape of AI technology. From deep expertise in large language models (LLMs), retrieval-augmented generation (RAG), and hybrid retrieval systems to fine-tuning techniques, agentic frameworks, and knowledge graph integration, I continuously refine my skill set to stay aligned with industry trends. My commitment to lifelong learning and active participation in cutting-edge research ensures I remain a valuable contributor to innovative AI-driven solutions that address real-world challenges.

Project Portfolio

Overview

At Fidelity Investments, I led transformative initiatives, including the consolidation of over 1 million business rules into fewer than 1,500 unique structures using AST, ASG, and ontology-driven approaches in Neo4j and RDF. I developed hybrid question-answering systems, achieving 94% accuracy, and designed a patented methodology for conditional QA generation, significantly enhancing efficiency and saving associates over 20 hours weekly.

As a Graduate Research Assistant at Northeastern University, I collaborated on groundbreaking projects, including building RAG systems with Qdrant to analyze 1,000+ SEC 10K reports and reducing manual analysis time by 95%. Additionally, I applied advanced statistical techniques to uncover insights in MLB draft datasets.

In personal projects, I have developed innovative solutions such as a melanoma detection system using CNNs, a financial sentiment classifier leveraging synthetic data, and a gesture recognition system with Transfer Learning, showcasing my expertise in deep learning, NLP, and predictive modeling. My diverse experience demonstrates a proven ability to deliver impactful, scalable solutions in AI and data science.

Work At Fidelity Investments

I am currently co-authoring research papers and filed patents based on my work at Fidelity Investments, where I developed a novel Guided Expert Thinking Process (distinct from Chain-of-Thought methods) capable of generating knowledge base triples with precise entity linking between subjects and objects. This innovative structure supports efficient node traversal using the TOG algorithm, setting a new benchmark for accuracy and reliability in knowledge retrieval. This experience highlights my ability to design and implement state-of-the-art prompt methodologies tailored to overcome specific challenges in LLM-based systems.

I contributed to the development of an Goal Oriented Agentic Framework that integrated LLMs to create a highly responsive and adaptive chatbot system. This involved assisting in designing an entity extraction pipeline to accurately identify user intents and key elements such as plan IDs and withdrawal types, achieving over 98% accuracy. I supported the implementation of a classification system to correctly identify query types, reducing misclassification errors by 35%. These accomplishments leveraged advanced NLP techniques and LLM capabilities to facilitate seamless, goal-oriented interactions, demonstrating expertise directly applicable to building persuasive, multi-session chatbot systems.

Led the ETL transformation of over 340k rules, consolidating them into fewer than 1,500 unique rules by designing common structures and redefining grammars. Developed advanced rule parsers using AST, ASG, and ANTLR to analyze, parse, and map expressions to ontologies, ensuring consistency across the rule engine. Designed an ontology in Protege and integrated it with RDF (Star Dog) and LPG (Neo4j) for efficient rule reasoning and execution. Streamlined operations by collapsing complex script-based rules, reducing rule complexity by over 80% and enhancing processing efficiency, resulting in significant time savings for associates.

Advertising Spend Analysis and Financial Insights Automation(Research)

Collaborated with professors from Northeastern University and the University of Texas to analyze advertising spend patterns across 1,000+ SEC 10K reports, focusing on sales impact correlation. Designed a RAG system with Qdrant vector database to process and index 10,000+ financial document pages, enabling efficient query processing. Built a question-answering system to extract financial metrics and correlate advertising spend with sales performance, reducing manual analysis time by 95%. Developed an automated pipeline to analyze advertising expenditures across 100+ company filings per hour, streamlining financial insights.

Competing Risk Analysis on MLB Draft Data (Research)

Demonstrated expertise in statistical analysis techniques, including survival analysis and competing risks models, to evaluate draft day factors impact on players MLB careers.

Utilized weighted least-squares regression and proportional hazards models to analyze complex competing risks data, showcasing strong analytical skills and problem-solving abilities.

Analyzed extensive MLB draft data (2012-2016) using advanced statistical methods and python programming, revealing significant predictors affecting players trajectories, including draft position, signing bonus, and player type.

We showcased our research findings at the NESSIS conference, held at Harvard University on September 23rd 2023.

Projects Links

Melenoma Detection Using CNN

The project focuses on melanoma detection using Convolutional Neural Networks (CNNs). Melanoma is a type of skin cancer that can be fatal if not detected and treated early.

Gesture Recognition Using Neural Networks

Leveraging neural networks and Transfer Learning, designed and implemented a robust system capable of accurately recognizing and interpreting five distinct user gestures.

Bankruptcy Predictor

Developed a bankruptcy prediction successfully streamlined complex financial data, demonstrating expertise in predictive modeling and quantitative analysis.

Startup Funding Analysis

Built and deployed a Python-based machine learning model for startup success prediction leveraging data from 57,000 startup records. Created models, including Random Forest with hyperparameter tuning and CAT Boost algorithms.

Research Article Summarizer

I frequently engage with research papers, but occasionally find them overwhelming or have spare time. As a solution, I developed a project: a straightforward article or text summarizer. This tool utilizes frequency analysis to identify essential sentences within a given text. Its purpose is to swiftly generate summaries for lengthy paragraphs or articles, simplifying the process of extracting key information..

Feel free to explore other exciting projects on my GitHub profile, all conveniently listed at the end of the page. For further discussions or collaborations, please don’t hesitate to contact me on LinkedIn. If you’re a recruiter interested in my qualifications, my detailed resume can also be found at the bottom of the page. Thank you!

Extracurricular Activities

I was a member of the National Cadet Corps (NCC) of India, where I engaged in various community development activities such as flood rescue operations in Visakhapatnam during Cyclone Hudhud in 2014, cleanliness programs, and participation in NCC Day activities, alongside my regular training activities.

Furthermore, I participated in various symposiums and other technical events in my college. As a part of an inter-college event, I visited different colleges to present my idea on weapon detection using image processing. This helped me to augment my planning and public-speaking skills.

During the annual techfest at RMD Engineering College, I also organized a national-level technical symposium event titled Saankethika 2K19 as a Joint Secretary.

Additionally, I led my college basketball team as a captain from my sophomore year and brought laurels to our institution. Participating in such varied activities and events helped strengthen my teamwork and organizational skills.

LINKEDIN

GITHUB

RESUME