presentation link : https://www.canva.com/design/DAGqJBKmv9g/ju7299B9RCCKX0yiTu1DUA/edit?fbclid=IwY2xjawK3vztleHRuA2FlbQIxMQABHu4dzzxIDnjV74BTisiqh8Bh_ENe50p7WdvJzCrmpfvzhReAOMN5hY7GSDgM_aem_WvpRXGvdrJtoWJ5vL2a8xw

AINS LegalHackers – AI Legal Assistant for All

Inspiration

In Tunisia, most citizens struggle to access clear, timely, and affordable legal help. Legal documents are often written in French or Modern Standard Arabic, while people speak in Darja (Tunisian dialect). Many live in rural areas, far from administrative offices, and lack internet access or smartphones. We wanted to change that.

What We Learned

We learned that democratizing legal access requires more than just AI — it demands cultural adaptation, low-tech accessibility, and community validation. We also deepened our skills in multimodal AI, language models, and Tunisian legal data.

How We Built It

Chatbot: Built with Mistral-Saba-24B (via Groq) and fallback to DialoGPT.
RAG: Integrated a FAISS-based retriever with Tunisian legal documents.
TTS/STT: Voice support using PlayAI TTS and LinTO ASR for Tunisian Arabic.
OCR: EasyOCR for scanning and analyzing official documents.
Frontend: React-based interface generated via Bolt AI.
Backend: FastAPI handling AI routing, voice, OCR, and chatbot APIs.
Multi-channel Support: Designed for SMS, USSD, and web.

Challenges We Faced

Finding or fine-tuning models that understand Darja fluently.
Running inference efficiently under low-resource conditions.
Bridging high-tech AI tools with low-tech devices like feature phones.
Creating real-time, explainable legal reasoning using AI.

Our Mission

We’re not just translating law. We’re making it understandable, accessible, and fair — for everyone in Tunisia, no matter where they live or what device they use.