Abdul Rafay

Built an intelligent document management system that automates verification workflows using RAG architecture.

I designed a multi-format ingestion layer with dedicated microservices for parsing PDFs, CSVs, XLSX, and DOCX files, extracting content into a normalized tree structure stored in the database. Each node in the tree carried a unique identifier, enabling granular updates, allowing users to modify any specific node without transmitting the entire document over the network, drastically reducing payload size and improving real-time collaboration. The tree structure also allowed efficient HTML reconstruction for rendering documents in the browser. Integrated vector embeddings for semantic search and applied prompt engineering techniques to minimize LLM hallucinations.

I built the full-stack application with React on the frontend and Node.js on the backend, connected to a vector database for context-aware document retrieval. The system replaced manual document review processes, significantly reducing processing time and human error.

Architected a full-stack platform connecting videographers with clients, managing end-to-end workflows including booking, payments, and digital asset delivery for 500+ monthly active users.

I built automated post-production pipelines using AWS MediaConvert for video processing, handling thumbnail generation and low-resolution rendering, and AWS Transcribe for generating transcriptions with 95% accuracy. I designed an event-driven architecture in which S3 uploads trigger Lambda functions to process assets automatically, eliminating manual intervention. I developed the client-facing interface in React and Next.js with a responsive design, integrated Stripe for payment handling, and built the backend API layer with Node.js and Express.