Pranav Rao

Computer Science graduate student skilled in machine learning and NLP, with hands-on experience in LLM-based sentiment analysis, model evaluation, and fine-tuning Hugging Face models. Demonstrated ability to communicate effectively through grading and feedback, enhancing student comprehension

• Put a Price on It! – Full Stack Social E-commerce Platform, 07/23-present, Full-stack web app for social e-commerce recommendations, enabling users to post items, like/unlike listings, search by item name or poster, and influence buying interest through community engagement. Developed a CRUD-enabled item catalog with liking, searching, and user-based interactions. Managed backend logic, user authentication, and integrated Nodemailer for email alerts. Deployed on Render with environment configs and CI/CD pipelines.
• Level Up Meds – Senior Capstone Project Medication Management App, 08/22-05/23, Medication management web app developed for a local start-up in a team of 5 developers. Developed a user-specific calendar for medication and appointment tracking. Implemented text-to-speech accessibility features for visually impaired users. Used Firebase for backend data and authentication.
• Image Express – Full Stack Machine Learning Application for ACS 560000, 09/25-12/25, Led a 3-person team to design and deploy a full-stack image classification system for character recognition across 2 distinct datasets (Pokémon, Thomas the Tank Engine). Developed a Flask backend for image uploads and inference, with a SolidJS and Tailwind frontend deployed on GitHub Pages. Created and fine-tuned TensorFlow/Keras models for both datasets, achieving 0.85 validation accuracy on the Thomas the Tank Engine dataset specifically. Coordinated development using JIRA, delegating tasks and ensuring timely progress across frontend, backend, and ML components. Made key architectural and implementation decisions, including converting models to TensorFlow Lite to meet deployment storage constraints, returning highest prediction and its corresponding label from backend, and proposing the basic architecture be server-client. Resolved critical development issues (e.g., Git/.gitignore conflicts, deployment limitations) and guided members in using Hugging Face CLI to navigate Hugging Face Hub.