← Back to Impact

Multilingual Agentic AI Workflow Hub

$2B Annual Savings80+ Countries60+ Languages

Context: Mercedes Benz Research & Development Bangalore – automotive domain.

Challenge: Disconnected multimodal applications (text-to-audio, video-to-video, doc-to-doc) across 80+ countries and 60+ languages. Needed a unified orchestrator with accent preservation and native tone.

Solution: Architected an MCP-based Agentic AI Orchestration Platform using LangGraph and LangChain. Integrated HuggingFace Coqui TTS, Whisper SST, and MarianMT for on-prem; Azure OpenAI and Cognitive Services for cloud. Built a Streamlit-based fallback for language mismatch detection. Created a dashboard measuring operational, functional, and non-functional metrics.

Outcome: Saved $2 billion annually. Serves stakeholders across 80+ countries with seamless data handoffs between independent AI modules.

Tech Stack:

Azure AI FoundryLangGraphLangChainMCPWhisper SSTCoqui TTSMarianMTStreamlitPythonMongoDB