The models most of the world uses are made by companies covered elsewhere or not covered at all, so what is built in India is different. It is India's own attempt at language models trained on Indian languages, led by Sarvam, voice companies like SoundHound AI, Sanas and Cartesia that chose Bengaluru for their engineering, and a robot model at Skild AI. Around them sits a group of young platform companies, many founded in India, like TrueFoundry, Concentrate AI and DATAmundi. Bengaluru holds nearly all of the work, and Lexsi Labs, a research company in Mumbai, gives that city a place too.
For an engineer, very little of this work is training a large model from scratch. Most of it is building with models and running the infrastructure underneath them, so the GenAI engineer is the central role here, more than anywhere else among product companies, with cloud and DevOps engineers running the GPU clusters and model platforms.
AI labs and model platforms is one of the Industry Verticals of
The Market Segments go from the model itself to the plumbing that turns it into a product.
Language models for Indian languages, voice and speech models, and a general model for robots come from companies like Sarvam, SoundHound AI, Sanas, Cartesia and Skild AI.
Training and serving, the gateway, the agent and the retrieval, and the data labelling and evaluation that judge a model come from companies like Anyscale, TrueFoundry, Lyzr AI, Surge AI and Galileo.
What AI labs and model platforms hire software engineers for, and which roles they hire, is on