🐪 Khaleeji-FinLLM-7B-Instruct

The World’s First Multi-Lingual Financial AI Agent for the Gulf Region.

Khaleeji-FinLLM is a specialized 7B parameter Large Language Model fine-tuned to navigate the complex financial, legal, and regulatory landscapes of the GCC (Saudi Arabia, UAE, etc.). It is designed to be the ultimate bilingual assistant for professionals and businesses in the Middle East.

Key Features

Bilingual Mastery: Seamlessly understands and responds in both high-level English and formal Financial Arabic. Regional Expertise: Specifically trained on UAE Corporate Tax laws, Saudi Vision 2030 FDI regulations, and Islamic Finance (Zakat calculations). Agentic Ready: Optimized for use in autonomous agent architectures (like LangGraph) with high reasoning accuracy for tool-calling. High Performance: Built on the Qwen2.5-7B architecture and fine-tuned on the AMD Developer Cloud using MI210 GPUs.

Use Cases Tax Advisory: Get instant answers on UAE Corporate Tax exemptions and Small Business Relief. Investment Guidance: Understand Saudi FDI conditions and foreign ownership laws under Vision 2030. Islamic Finance: Perform complex Zakat calculations based on current Sharia-compliant standards. Market Analysis: Analyze bilingual financial reports and regional news with localized context.

Quick Start

Run this model locally with Ollama:

ollama run muhammadanique81/Khaleeji-FinLLM-7B-Instruct

Training Details

Dataset: 500k+ curated bilingual financial samples from the Gulf region. Hardware: Trained on AMD MI210 GPUs (AMD Developer Cloud) using the ROCm ecosystem. Method: LoRA (Low-Rank Adaptation) fine-tuning followed by 8-bit quantization for optimal local performance.

🔗 Project Links

Live Demo: khaleeji-ai.vercel.app

Hugging Face Weights: huggingface.co/anique-1/khaleeji-qwen2.5-7b-finllm

Built by Muhammad Anique for the LabLab AI & AMD Hackathon.