I've spent the last year building AI tools and full-stack apps. Here's what I have so far

I'm Sufiyan, a developer from Satara, India. Most of what I build sits somewhere between a normal web app and an AI feature that has to work in the real world.

The one I'm most proud of is Groundwork. You upload your own documents, ask questions, and get answers that point back to where they came from. It searches by keyword and by meaning, then re-ranks the results before writing the answer. At first it kept losing its search index because of free hosting, so I moved the storage to Qdrant Cloud. I also tested it with Ragas and got 1.00 on faithfulness and 0.71 on answer relevancy. The second number is lower than I'd like, and that's what I want to improve next.

LeadAI came from a simple problem. When a lead comes in and nobody replies quickly, that lead usually goes cold. So an AI agent writes the follow-up, sends it by email, and logs the lead in a Google Sheet. It runs on n8n.

AI CodeReview Pro is a set of agents that reviews code changes. Anything risky, like auth or payments, gets passed to a person instead of being approved automatically.

NotesToReport is open source. It turns raw notes into a report with citations, and it blocks the report if the claims don't match the notes (faithfulness under 0.85).

I also built Aurelia Retreat, a booking site with an AI assistant and a lead dashboard.

All of them have live demos and code on my portfolio, along with how each one works:
https://sufiyankhan-dev.vercel.app

I'm taking on a few freelance projects now, mostly AI automation and full-stack builds. If you have something like this in mind, comment or message me.

And if you've shipped a RAG app, what did you use to measure whether the answers were good?