Responding to an RFP means re-answering questions the company has already answered a dozen times, hunting through old proposals for the wording that worked. It is slow, and the risk clauses that matter are easy to miss under deadline.
What I did
01Ingestion pipeline that reads long RFP documents at over a thousand pages per minute.
02A LangGraph agent retrieves matching prior content and drafts responses section by section rather than as one undifferentiated block.
03Risk clauses are surfaced for human review, and output is generated as DOCX so it drops into the existing approval process.
04A human approval gate sits before anything leaves the system.
Outcome
Proposal preparation time cut by roughly 70%, with the reviewer spending their attention on judgement calls instead of retyping.
