Shanay Ghag
I am a Data Scientist at Great American Insurance Group, optimising insurance processes using AI & ML. I received my M.S. in Computer Science from University of Southern California, and B.E. in Information Technology from Pune Institute of Computer Technology.
My research focuses on efficient fine-tuning algorithms for multimodal large language models, inference optimisation methods and systems for ML
Experience
My journey in research and industry
Work Experience
Industry roles and internships
Data Scientist
Great American Insurance Group
Developing tools for insurance automation using Machine Learning, Deep Learning, NLP, and Agentic AI. Specializing in deep research agents, fraud detection systems, and multi-modal llm fine-tuning
Machine Learning Engineer Intern
Ardent Privacy
Engineered ML/NLP pipeline for automated data privacy, achieving 80% faster sensitive data discovery and 91% classification accuracy using ensemble methods (Char-LSTM + XGBoost) deployed on AWS
Data Science Intern
Veritas Technologies LLC
Built Dense Passage Retriever with GraphCodeBert for intelligent code search (MRR: 0.272) and developed retrieval-augmented algorithms for automated code generation from agile user stories
Data Science Intern
FinSoftAI Solutions
Developed real-time NLP engine for financial sentiment analysis across news and social media, supporting investment research, ESG scoring, and trading decisions using Target Dependent Sentiment Analysis
Data Science Intern
iMocha
Enhanced AI-EnglishPro assessment tool by 13% accuracy through advanced sentiment analysis and grammar detection. Built internal code similarity detection tool for developer evaluation platform
Research Experience
Academic research
Research Intern
University of South Carolina
Working on a parameter-efficient finetuning algorithm using curvature-aware optimization and neural reprojection for faster convergence with fewer parameters
Research Assistant
University of Southern California - Institute of Creative Technologies
Worked on a contrastive learning-based model for ensuring dialog consistency in negotiation dialog systems
Research Assistant
Pune Institute of Computer Technology - Computational Linguistics Lab
Researched algorithms for generating Bloom's Taxonomy-aligned questions and generating low-level code from natural language
Projects
A selection of my recent work in machine learning, data science, and AI research
Geometric Reprojection Instruction Tuning (GRIT)
A parameter-efficient fine-tuning framework that combines LoRA with curvature-aware optimization and neural reprojection for efficient adaptation of vision-language models. Research work advised by Dr. Amitava Das
Speculative Decoding
Speculative Decoding implementation based on 'Accelerating Large Language Model Decoding with Speculative Sampling' paper by DeepMind, 2023
Genetic Algorithm for TSP
Advanced Genetic Algorithm for solving the Traveling Salesman Problem with k-nearest neighbor initialization, 2-opt mutation, and multi-tier breeding strategies
Get In Touch
I'm always open to discussing new projects, research opportunities, or collaborations. Feel free to reach out through any of the channels below.