Divash Upadhyay

Built and maintained AI.SEE, an on-premise AI visual inspection platform running on 25+ factory edge machines. Rebuilt the operator inspection UI as an industrial HMI using React 18, TypeScript, and Vite, including server-side filtering/sorting, inspection controls, and an Expert Export that generates TP/FP/FN/TN Excel reports. Developed backend features with NestJS, TypeORM, ZeroMQ, and Socket.IO: export pipeline, real-time sequence progress/stop events, Azure Blob Storage integration, storage/volume management, and authentication hardening. Implemented a fleet monitoring console in Next.js/React 19 with scheduled health checks via SSH (CPU/RAM/GPU/network) and reporting views. Debugged and fixed production issues across Linux edge pipelines (camera discovery/order, upload race conditions, Azure durability, demosaic bug, AI fallback), verified on live hardware; managed Docker Compose deployments and hot patching in customer environments via SSH/RDP/KVM and private registry operations.

Led migration of a legacy JavaScript codebase to TypeScript across frontend and backend, improving reliability via strict typing and automated tests. Architected microservices using NestJS and Express over PostgreSQL and MongoDB. Built BullMQ/Redis job queues with idempotency keys and batched processing to prevent double processing on retries.

Built REST APIs with NestJS/Express for an e-commerce platform supporting 10,000+ daily users, achieving p95 query time under 150 ms via PostgreSQL indexing and query tuning. Improved peak throughput by ~3.5x using Redis caching on heavy read paths. Set up Docker-based deployments and GitHub Actions CI/CD to cut deployment time in half. Integrated FFmpeg into Node.js services for automated video transcoding, thumbnail generation, and compression.