Mrinal Sharma • Mrinal Sharma •
Full Stack Engineer • Full Stack Engineer •
I build interactive experiences that blend code, design, and artificial intelligence.
Scroll
About Me
I am a Master's Candidate in Computer Science & Engineering at the University of Notre Dame (Class of 2026, 4.0 GPA), and hold a B.Tech in CS from VIT (3.88 GPA).
I specialize in building scalable, high-performance applications at the intersection of Full Stack Engineering and Generative AI. My passion lies in creating robust architectures for AI-driven solutions that solve real-world problems.
With experience at RecruitMate and as a Graduate Teaching Assistant, I have honed my skills in Next.js, Cloud Computing (AWS, Kubernetes), and modern AI integrations like Gemini and Vercel AI SDK.

My Experience
Software Engineering Intern
Aug 2024 - May 2025RecruitMate
- Improved Time to Interactive by 380ms (2.1s → 1.72s) via Next.js Server Components.
- Designed Prisma schema for 40+ universities maintaining sub-50ms queries.
- Optimized LCP reducing bundle size by 28%.
Graduate Teaching Assistant
Aug 2024 - Dec 2025University of Notre Dame
- Directed weekly technical labs for 50+ students in Data Structures & Algorithms.
- Conducted code reviews for C++ and Java implementations of Hash Maps, Trees, and Graphs.
Selected Works
- Startup platform for student-athlete recruitment facilitating video portfolios.
- Improved TTI by 380ms (2.1s → 1.72s) via Next.js Server Components.
- Optimized LCP by 28% reducing bundle size significantly.
- Designed scalable Prisma schema for 40+ universities.
Next.jsTypeScriptPrismaTailwind

- AI-Powered Car Rental Platform for peer-to-peer sharing.
- Reduced image load latency by 60% via AWS S3 + CloudFront CDN.
- Built automated damage detection using Google Gemini Multimodal Vision.
- Secured payments via Stripe Connect + NextAuth.js.
Next.jsAWS S3CloudFrontGeminiAIStripe

- Reduced database write load by 80% via Redis write-behind caching, persisting to PostgreSQL every 60s or on session end.
- Achieved <50ms end-to-end typing latency by building a custom WebSocket broadcast server, supporting 100+ concurrent rooms on a single node.
- Accelerated dev cycle by 30% through Turborepo monorepo with shared types between Next.js client and Express backend.
Next.jsNode.jsWebSocketsMonaco Editor

© 2026 Mrinal Sharma • Neo Banana V2

