Umang Agarwal

Developed Roverpod Genie, an AI meeting assistant that autonomously joins virtual meetings, speaks with < 300ms latency, records, streams live to millions, and delivers real-time contextual intelligence. Built a multimodal RAG system over enterprise knowledge bases, alongside scalable Dockerized browser environments and an HLS streaming pipeline to support low-latency, high-concurrency video proce

Designed and developed a cross-platform badminton court booking app using React Native and Expo, streamlining scheduling and improving facility utilization. Integrated React Query, Expo SQLite, and push notifications for real-time data fetching, offline support, optimistic UI updates, and instant booking alerts.

Built an AI-powered photo-sharing platform that uses real-time facial recognition and UPI integration to deliver personalized event photos via WhatsApp, automating photographers’ workflow end-to-end. Engineered high-speed face matching infrastructure using FaceNet embeddings, Pinecone, and self-hosted ML inference servers to ensure low-latency, scalable photo retrieval across large datasets.

Architected and developed a privacy-first SaaS platform that enables users to monetize personal data while maintaining full ownership and 100% compliance with GDPR, CCPA, and RTI regulations. Built secure, scalable infrastructure including OAuth-based data retrieval from 30+ third-party services, encrypted AWS S3 data vaults, and a distributed processing pipeline using AWS Lambda, SQS, EC2, and S3

Worked on a payout prediction model for delivery executives, reducing manual workload for the Data Science team and contributing to a projected 17% reduction in cost per delivery. Implemented Kafka-based batch processing and instrumented key gRPC APIs, ensuring low-latency performance during peak events and enabling proactive outage prevention.

Completed a Bachelor of Engineering with Honors in Computer Science. Gained a strong foundation in computer science principles and practices.