Hi, I'm Twisha Shah
MS Computer Science Student | Agentic AI & LLM Specialist
I build |
Currently pursuing my Master's in Computer Science at Georgia Tech, specializing in Machine Learning. This summer, I interned at Qualcomm building Agentic AI pipelines, and prior to that, I spent two years at JP Morgan as an SDE. I'm passionate about Large Language Models, Agentic workflows, and building intelligent AI systems that solve real problems. Always excited to learn, collaborate, and create impactful solutions.
Twisha Shah
Software Engineer.
Curious generalist. Focused builder.
Education
Master of Science in Computer Science
Georgia Institute of Technology
GPA: 4.0/4.0
Focus: Machine Learning and Systems
Relevant Coursework: Machine Learning, Conversational AI, Systems for ML, GPU Programming, Hardware Software Co-design for ML, Computer Vision, Data Analytics
Bachelor of Technology in Electronics Engineering
VJTI, Mumbai
GPA: 9.71/10
Focus: Computer Engineering
Relevant Coursework: Data Structures, Algorithms, Database Systems, Software Engineering, Computer Networks, Embedded Systems, NLP, Data Science, Image Processing
Experience
Research Intern
Madabhushi Lab, Emory University
- Contributing to the design and development of an Agentic AI framework for oncology, focusing on autonomous agents for hypothesis generation and multi-modal data curation.
- Developing autonomous agents and a collaboration framework to ingest and preprocess histopathology images and clinical metadata from multi-institutional cohorts for downstream AI analysis.
Applied ML/GenAI Intern
Qualcomm
- Architected a multi-agent MCP-based system with AI-driven orchestration and Pydantic validation to enable reliable, reproducible engineering workflows for data retrieval, model generation, simulation execution, and metric calculation on EDA results (e.g.: Ansys).
- Reduced LLM hallucinations by 90% by integrating database-driven dataframes across AI tools to extract accurate Snapdragon chip data.
- Enabled 100% of new hires to independently run simulations within 2 weeks (vs. 6 weeks before) and cut manual EDA and inhouse tools setup time from 2 hours to 15 minutes.
- Developed a two-layer MCP client-server framework with an intelligent Orchestrator for query parsing, permission control, and dynamic routing - achieving 30% faster chip design life-cycles.
Software Developer I
JP Morgan Chase & Co.
- Worked with a cross-functional team on delivering high-impact business features for the Equities OMS.
- Proficiently crafted MySQL queries for database operations & data retrieval, enhancing application efficiency.
- Spearheaded planning, decision-making, and execution to migrate Sybase to AWS-RDS database with a global team.
- Reduced operational costs by 50% and enhanced system reliability by 35%.
- Developed a JMX trigger to analyze and remove inactive caches, resulting in up to 45% reduction in memory usage.
- Developed Python script to automate DDL with Liquibase generation saving 4 weeks of manual effort.
Data Science Intern
Fractal Analytics
- Developed a reinforcement learning model to optimize ad recommendations for a website.
- Designed and implemented stochastic strategies with batch processing.
- Increased cumulative reward by 30% compared to A/B testing.
- Constructed a simulation model to generate synthetic data, accurately mimicking real-life web traffic scenarios.
Featured Projects
Soccer Strategy Optimization
Applied machine learning to cluster La Liga teams by playing style and predict match outcomes. Utilized K-Means clustering, Hierarchical Clustering, and PCA to analyze player performance and optimize strategy. Performed extensive feature engineering and data cleaning, removing outliers and normalizing statistical attributes for consistency across matches.
Applied K-Means clustering with StandardScaler-based normalization to group teams by playstyle and identify performance patterns. Evaluated multiple cluster sizes and optimized model selection using silhouette scores, achieving clear separation of team profiles for downstream predictive modeling.
Developed predictive models on engineered features to forecast match outcomes, comparing Logistic Regression, Random Forest, and SVM, and selecting the best-performing model based on cross-validation accuracy and F1-score.
AI Self-Checkout System
Led a team of 4 to build a real-time customer tracking app inspired by Amazon Go, leveraging YOLOv3, DeepSORT, face recognition, and socket communication for dynamic object tracking and UID mapping.
Skills & Technologies
Programming & Data
AI/ML & GenAI
Cloud & Infrastructure
Development & Tools
Management & Collaboration
Certifications
Achievements & Extra-curriculars
Hackathons
Code For Good (CFG'21) - 1st prize | Electrothon4.0 - 3rd prize
Panelist
Panelist at MTC Engineering event (JPMC) | Core Member, AWS DeepRacer Mumbai Chapter (JPMC)
Subject Matter Expert
Subject Matter Expert - CFG'23 Hackathon
Software Developer
Software Developer - Force For Good Hackathon (8 months)
Let's Connect!
I'm always excited about new opportunities in Agentic AI, Multi-modal AI, LLMs, and AI systems in the cloud. I love hackathons and building innovative AI projects. Feel free to reach out if you want to collaborate, discuss research, hire me, or just say hello!