Muhammad Zohair B.

With a Master’s in Data Science and 6+ years building AI systems, I build Document AI workflows for teams working with PDFs, invoices, medical and legal records, forms, reports, drawings, and private knowledge bases, using RAG, OCR, extraction, citations, and validation. I solve problems for the following industries: 𝗙𝗶𝗻𝗮𝗻𝗰𝗲, 𝗹𝗲𝗻𝗱𝗶𝗻𝗴 𝗮𝗻𝗱 𝗰𝗼𝗺𝗽𝗹𝗶𝗮𝗻𝗰𝗲 - KYC packets, loan files, bank statements, onboarding docs, invoices, source-of-funds records, and audit trails. I make sure to keep the data complete, validated, traceable and safe to move it into the next decision. 𝗟𝗲𝗴𝗮𝗹, 𝗰𝗹𝗮𝗶𝗺𝘀 𝗮𝗻𝗱 𝗰𝗼𝗻𝘁𝗿𝗮𝗰𝘁𝘀 - I handle matter files, contracts, exhibits, claim packets, legal and medical records, and long supporting document set, and build source-backed review workflows where every extracted fact can be traced back to the original record. 𝗛𝗲𝗮𝗹𝘁𝗵𝗰𝗮𝗿𝗲 - SOAP notes, coding support, prior-auth packets, claims documents, insurance forms, referrals, and outputs ready for EHR integration to reduce the documentation load on admin teams. 𝗖𝗼𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝗼𝗻 - plan sets, specs, schedules, RFIs, submittals, change orders, and scope docs, by building the automation in a way that it understands sheets, callouts, symbols, notes, schedules, and specs, and keep the answer tied to the right source. 𝗠𝗮𝗻𝘂𝗳𝗮𝗰𝘁𝘂𝗿𝗶𝗻𝗴/𝗹𝗼𝗴𝗶𝘀𝘁𝗶𝗰𝘀 - I focus on bills of lading, PODs, invoices, packing slips, supplier docs, certificates, delivery notes, and quality records, while ensuring sensitive parts like bad scans, missing signatures, and mismatched fields are solved before the document moves forward. Across these industries, I've built document workflows that turned things like plan sets, clinical notes, legal records, and regulatory files into cited answers, structured outputs and downstream documents. For example, my recent work includes achieving 98.8% cited answer accuracy on construction plan Q&A, 96%+ extraction accuracy on takeoff workflows, and reduced the review time from hours to minutes in compliance and legal document workflows, saving the teams from 25-30+ hours of manual review every week. If this sounds close to what you need, then let's schedule a 30 min call for an architectural review first. Stack: 𝗘𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻: PyMuPDF, PyPDF2, Tesseract OCR, AWS Textract, Azure Document Intelligence, Google Document AI, OpenCV, YOLO, schema-based JSON extraction 𝗥𝗔𝗚: LangGraph, LangChain, LlamaIndex, Pinecone, Qdrant, pgvector, Chroma, Weaviate, VoyageAI embeddings, metadata filtering, citation retrieval 𝗟𝗟𝗠𝘀: OpenAI, Claude, Gemini, Azure OpenAI, structured outputs, tool calling, evaluation sets 𝗕𝗮𝗰𝗸𝗲𝗻𝗱 𝗮𝗻𝗱 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄: Python, FastAPI, Node.js, Next.js, PostgreSQL, MongoDB, Redis, SharePoint, Google Drive, Power Automate, n8n, Make, Docker, AWS, Azure