Khaleeji AI is a professional-grade AI ecosystem designed to solve the critical gap in localized financial intelligence for the Middle East. While general models struggle with regional legal nuances and formal financial Arabic, Khaleeji AI provides high-precision advisory for the GCC market. The Technical Build: Custom Fine-Tuning: We custom-trained a 7B parameter LLM (Khaleeji-FinLLM) using LoRA/PEFT on a curated 500k+ token dataset. AMD Powered: The model was trained for 16 continuous hours on the AMD Developer Cloud using MI210 GPUs and the ROCm ecosystem. Agentic Intelligence: Built with LangGraph, the system isn't just a chatbotβit's an autonomous agent. It uses real-time web search (Tavily) to verify current market laws before responding. Deployment: Optimized with 8-bit GGUF quantization for low-latency local inference and available globally on the Ollama Hub.
Category tags:"Solving key issues around formal financial documentation should increase trust in AI over time. The AMD powered model built with LangGraph is a solid concept. Application of Technology: 5 rockets πππππ- The implementation of custom-trained LLM solves key issues with Arabic. Presentation: 4 rockets ππππ -The language is easy to understand and key components have been articulated well. Business Value: 4 rockets ππππ - This project could unlock value for the GCC market. Originality: 5 rockets πππππ - Creating high-precision for existing advisory really builds something new."
Sanem Avcil
"Presentation and demonstrations are clear. Technical information provided "
Jayashree H V