RAG & knowledge systems.

Answers grounded in your own documents, with the retrieval measured.

Start a project How I build it
  • RAG
  • FAISS
  • ChromaDB
  • Hugging Face

The problem

Most wrong answers from a RAG system are retrieval failures. The right passage never reached the model, so it filled the gap with something that sounded right.

Here’s how I build it

How I build it.

5 steps from the first conversation to launch. The diagram fills in as each part gets built.

  1. Step 01

    Collect the sources

    Every source gets a way in: PDFs, web pages, exports, images. Each document is split into chunks that keep their meaning, with the heading and context attached.

  2. Step 02

    Build the index

    Chunks become embeddings in a vector store, FAISS or ChromaDB depending on size and hosting. Keyword search runs beside it for exact names, codes and numbers.

  3. Step 03

    Find the right passages

    A question pulls back the passages most likely to answer it. With multimodal retrieval, that includes images as well as text.

  4. Step 04

    Answer from the sources

    The model answers from what was retrieved and says which document each part came from. If nothing relevant comes back, it should say so.

  5. Step 05

    Measure the retrieval

    A test set of questions checks whether the right passages come back. The score shows what to fix: the chunking, the index or the way questions are handled.

What you get.

All of it, or only the part you need. Each piece can be scoped on its own.

Built with

  • RAG
  • FAISS
  • ChromaDB
  • Hugging Face
  1. 01Ingestion and chunking pipeline for your corpus
  2. 02Vector store selection, indexing and hybrid retrieval
  3. 03Multimodal retrieval across text and images
  4. 04Retrieval scoring, so you can see what it actually found

Where I’ve done it.

Quoted from my CV, then the builds you can open and check for yourself.

  • Padelos

    Associate AI Engineer

    Jan 2026 — Present

    • Manage agent memory with vector stores (FAISS, ChromaDB) for retrieval, and Redis and PostgreSQL for short-term and persistent state.
  • UH

    University of Haripur

    BS Computer Science — CGPA 3.79 / 4.0

    Sep 2021 — Jul 2025

    • Built a multimodal RAG system combining text and image understanding with LangChain, FAISS/ChromaDB and Hugging Face Transformers, served through a FastAPI backend for quiz generation, Q&A and mock-interview simulations.

Builds.

All projects

Questions clients ask.

Not covered here? Ask me directly.

What documents can it work with?

Most text formats: PDFs, Word files, web pages and exports from your tools. Scanned pages need their text extracted first, and multimodal retrieval can search images directly.

Will it make things up?

It can, which is why the retrieval is measured and each answer cites its sources. When the right passage isn’t found, the system should say so rather than guess.

Can our data stay on our own servers?

Yes. FAISS and ChromaDB both run on your infrastructure, and an open model from Hugging Face can run there too if the documents can’t leave it.

How does it stay up to date?

The ingestion pipeline runs again when documents change, so new and edited files are indexed without rebuilding everything.

Other services.

Work with me

Have a project like this?

Tell me what you want to build, and I'll tell you how I would approach it.

Let’s talk