At Al–RafayConsulting, I implemented a multi-stage AI document classification project using text-based and vision-based LLMs. I designed the architecture in Python and FastAPI to process files from Azure Blob and SharePoint Cloud.
I improved file processing speed by ~90% by maintaining a persistent connection across files, and implemented document sub-classification using Mistral OCR LLM. I also collaborated with a senior team member to integrate the AI service with a frontend application.
In my projects, I built a multi-agent research system and a LangGraph college assistant that routes questions across academic handbook, fee structure, and general chat sources. I also developed and evaluated a tweet sentiment classifier, investigating data leakage before retraining it.
