Government-backed LLM and multimodal AI for Indian languages
BharatGen is a government-funded AI consortium under IIT Bombay building large language models and multimodal foundation models optimized for Indian languages. The tech stack emphasizes LLM inference and orchestration (LangGraph, DSPy, vLLM, Ray, SLURM) paired with evaluation infrastructure (MLflow, Weights & Biases, DVC), reflecting a core focus on model quality and operational stability. Current project work centers on evaluation pipelines, RAG systems, and agentic architectures — suggesting the organization is moving beyond foundational model training toward production-grade deployment and autonomous agent systems.
BharatGen operates as a government-funded nonprofit consortium under the Technology Innovation Hub at IIT Bombay, with backing from India's Department of Science and Technology. The mandate spans developing efficient LLMs and multimodal models for Indian languages, building multilingual data repositories, fostering public-private partnerships, and strengthening India's AI talent ecosystem. The organization is structured around engineering-focused work, with active hiring concentrated in senior and mid-level technical roles across India. Current work includes foundational model development, RAG pipeline implementation, and agentic AI system architecture, alongside critical work on AI safety, trustworthiness, and evaluation metrics.
Primary stack includes LangGraph, DSPy, AutoGen, and CrewAI for model orchestration; PyTorch and TensorFlow for training; vLLM, Ray, and SLURM for inference and scaling; MLflow, Weights & Biases, and DVC for experiment tracking and evaluation.
Active projects include foundational generative AI model development, AI evaluation pipeline creation, retrieval-augmented generation systems, agentic AI architectures, and CI/CD release gate implementation for production deployment.
Other companies in the same industry, closest in size