AI/ML engineer building local-first, CPU-efficient ML tools — from model internals to usable products.

github

Yash Agrawal  • Peerlist

Real-time AI tool that auto-crops 16:9 livestreams to 9:16 by tracking faces + objects — runs on CPU, zero GPU needed.

-AI Smart Video Resizer

🗣️ Built a Hinglish translation tool with FastAPI backend and React frontend, delivering accurate translations and increasing daily user interactions by ~25%. 🚀 Deployed using CI/CD (Render backend, Vercel frontend), reducing deployment time by ~50% and ensuring modular, maintainable architecture.

Bolbhasha

Find any file using plain English using intent. Built with sentence embeddings + Qdrant vector DB. Docker-ready.

-EdgeFinder – Semantic File Search Engine

• Designed a web application that increased GitHub repository preview efficiency by 90%, eliminating the need for code downloads and enhancing user experience. • Gained deep knowledge of advanced tech tools proving flexible and creative in tackling issues. This led to a 30% cut in the time needed to develop projects.

InstantRepo

Co-built RL environment (OpenEnv Hackathon 2026) where AI agents resolve non-performing bank loans under RBI rules.

-LoanRestructureEnv – RL Fintech Environment

• Designed a desktop application that tracks daily screen time and app usage, resetting data every 24 hours. • Helps users monitor productivity and reduce unnecessary screen time, improving time management habits.

Windows Digital Wellbeing App

• Built an engaging mobile app featuring a “Surprise Me” button that generates witty definitions for common terms. • Integrated share and save functionalities, increasing user engagement and social media interactions.