Pranav Chopra

LeakSight is a cybersecurity research project that uses machine learning to visualize and analyze software-based side-channel leaks. These are small timing or system behavior patterns that can unintentionally reveal information. The project builds a safe and controlled environment where two programs run on the same computer to show how these leaks appear and how simple defensive techniques can reduce them. By combining data analysis, machine learning models, and interactive visual dashboards, LeakSight turns complex security ideas into clear visuals that help people understand and prevent side-channel vulnerabilities.

Built and integrated a TensorFlow/Keras CNN audio classifier into a vocal-coaching web app backend, delivering real-time feedback on pitch, timing, vibrato, and clarity. Achieved 99% validation accuracy and 98% test accuracy on a held-out dataset, with an end-to-end audio preprocessing and batching pipeline for low-latency inference. Designed for rollout to 300+ musicians at the Canadian Musicians Co-operative, enabling scalable, real-time coaching across the cohort.

Recently designed and launched a user-centric CPA website that boosted business by 20% through clear information architecture, fast load times, and streamlined contact/lead flows. Built a full-stack app with React + Next.js (routing, SSR, API routes), TypeScript, and Tailwind CSS—delivering a fast, component-driven UI and clean developer ergonomics. Implemented Node.js/Express REST APIs and Next.js API endpoints (email via Nodemailer), integrated external APIs, and deployed on Firebase/Google Cloud with secure env secrets and production monitoring.