Atlas Industries
A bilingual (EN/AR) agentic RAG knowledge assistant on an 8-node LangGraph workflow — 90% routing accuracy and 0.91 answer relevancy on a DeepEval suite.
Overview
Atlas Industries is a bilingual (English/Arabic) enterprise knowledge assistant — my capstone for the Sprints AI & ML Bootcamp (U.S. Embassy / American Center Cairo). It runs an 8-node LangGraph agentic workflow over Groq’s Llama-3.3-70B, with BGE-M3 multilingual embeddings and FAISS vector search behind an API layer.
I owned the data ingestion and vector-store pipeline end to end — multi-format EN/AR parsing, chunking, embedding, and indexing — and built the evaluation tooling that validated the system.
What I built
Agentic RAG workflow
An 8-node LangGraph graph routing queries across domains — a reusable GenAI framework serving retrieval through an API.
Multilingual retrieval
Owned ingestion of multi-format EN/AR documents into a FAISS index using BGE-M3 embeddings.
Measured quality
Reached 90% routing accuracy and 0.91 answer relevancy, validated with a 5-metric DeepEval suite over gold test cases.
Built with
- LangGraph
- Groq Llama-3.3-70B
- BGE-M3
- FAISS
- FastAPI
- DeepEval
Have a project in mind?
Whether it’s a business website, an app, or something AI-powered — tell me about it and I’ll reply the same day.