I'm Abdelmalek, a full-stack developer from Algeria. Today I'm launching Atlas Bac — a study assistant for Algerian baccalaureate students studying History and Geography. The idea came from watching how much students struggle during exam season — the syllabus is dense, revision time is short, and generic AI chatbots don't actually know the specific curriculum they're being tested on. So I built Atlas Bac as a RAG-powered assistant: it retrieves content directly from the real bac curriculum before generating an answer, instead of relying purely on general AI knowledge. The goal is answers that are actually grounded in what students need to know, not just plausible-sounding text. A few technical bits I'm proud of: 🔹 Hybrid retrieval (metadata filtering + semantic search) to handle exam-style questions 🔹 Query decomposition for compound questions 🔹 Structured tables rendered outside the LLM entirely, to cut hallucinations on fact-heavy content Stack: Django (Channels/ASGI, Celery, Redis), Next.js, PostgreSQL + pgvector, Gemini. This is still early-stage, and I'd genuinely love your feedback — especially from anyone who's built RAG systems for non-English or RTL content, since that's been one of the trickier parts. Thanks for checking it out! 🙏

مرحبا بكم في اطلس البكالوريا

AI study assistant for Algerian bac history & geography

AI study assistant for Algerian bac history & geography