Building Mahindra’s VehicleGPT on MongoDB Atlas

VehicleGPT, developed by Mahindra & Mahindra, is an AI-powered assistant integrated into their mobile app to enhance the ownership experience by providing answers to natural-language questions about vehicle features, controls, warning symbols, and manual content. It is notable for being deployed across all passenger vehicle brands and trim variants in Mahindra’s portfolio, offering a comprehensive digital support system within the app drivers already use. The system utilizes MongoDB Atlas for data management and Voyage AI’s multimodal-3 model for embeddings, while generating responses with GPT-5.1, and has successfully handled over 15,000 customer queries. VehicleGPT's architecture is designed to streamline retrieval by converting complex manual content into MongoDB documents, enabling efficient hybrid search that combines semantic and exact-match retrieval. This approach eliminates the need for a separate vector database and custom hybrid search code, reducing complexity and cost while maintaining flexibility for future expansions, such as multilingual content. Mahindra's strategic decision to embed this assistant within the app signifies a broader initiative to continually improve the ownership experience post-purchase, setting a precedent in the automotive industry for integrating AI with vehicle ownership journeys.