Tarun Sharma

Built and deployed a real-time single-trial P300 EEG classifier for forensic use using a multi-branch CNN with ERP features and xDAWN spatial filtering. Designed subject-adaptation pipelines and deployment-grade evaluation/audit logging, improving cross-subject performance using ~60 calibration trials and pushing G-Mean from the low 0.70s toward acceptable sensitivity/specificity tradeoffs.

Developed a production-grade Redis-style in-memory datastore in Go with full RESP support, handling 10,000+ concurrent connections with 99.9% uptime and ~1ms latency. Built an AI chatbot (95% response accuracy across 1,000+ daily interactions) and a CV-based ID scanning module (95% extraction for 500+ users), plus end-to-end ML/data analytics for 5+ client datasets achieving 90%+ accuracy.

Engineered a discussion board feature using Node.js and MongoDB, increasing student engagement by 30%. Optimized backend APIs via efficient schema design to reduce latency by 20%, and conducted mobile beta testing achieving 95% crash-free performance on Google Play.

Built a predictive maintenance system using Java/Spring Boot and ML algorithms, cutting service delays by 25%. Refined a microservices architecture to improve API response times by 35% and maintained a full-stack platform (React.js, Go) serving 1,000+ monthly active users.

Facilitated learning for 200+ students for Data Structures & Algorithms, improving exam performance by 15% through weekly recitations. Organized office hours and tutoring sessions to provide targeted guidance for assignments and exam preparation.

Earned a Bachelor of Science in Computer Science from Stony Brook University, graduating in 2023.