Government Tech · Government: Roads & Transport Authority (UAE)
RTA Mahboob: Multilingual RAG & Text-to-SQL Platform
Bilingual EN/AR RAG for 1000+ UAE government users with text-to-SQL over 500+ tables, cutting latency 73%, from 15s to 4s.
- Role
- AI Engineer
- Built multilingual RAG system with 73% latency optimization for a national transport authority.
- Industry
- Government: Roads & Transport Authority (UAE)
- Timeline
- Jul 2025 - Sep 2025
Business impact
- response time
- 4s
- latency improvement
- 73%
- accuracy
- 95%
- active users
- 1000+
- database tables
- 500+
- languages
- English + Arabic
Overview
Built RTA Mahboob for the UAE Roads and Transport Authority: a multilingual RAG platform that routes every query by language into a dedicated English or Arabic pipeline for low latency, alongside text-to-SQL over 500+ database tables exposed through FastAPI.
Language-aware routing removed the cross-lingual translation hop that dominated response time, cutting latency from 15 seconds to 4 seconds (73%) while holding 95% answer accuracy for 1000+ active government users.
Technical architecture
- AI/LLM
- GPT 4o with LangChain and LangGraph
- RAG
- Azure AI Search with multilingual embeddings
- Database
- PostgreSQL with 500+ tables, schema vectorization
- Backend
- FastAPI with Docker
Key features
- Multilingual RAG System (English + Arabic)
- Optimized retrieval across 120+ public transport services
- Text to SQL across 500+ PostgreSQL tables
- Schema vectorization for intelligent SQL generation
- Intelligent Document Drafting workflows
Challenges solved
- Optimizing response latency from 15s to 4s (73% improvement)
- Building multilingual embeddings for English and Arabic
- Schema documentation for 500+ tables
Tech stack
- LangChain
- LangGraph
- Azure AI Foundry
- Docker
- FastAPI
- RAG
- PostgreSQL
- Azure AI Search
- GPT 4o
- Python