What AI labs and model platforms hire for

Industry Vertical 9, AI labs and model platforms, in Domain III, Data platforms and AI.

The short answer

AI labs and model platforms build language, speech and robot models, and the tools that train, serve, connect and test them. Very few of them train large models from scratch in India. What they mostly hire for here is building with models and running the infrastructure underneath. So the GenAI engineer is the central role, far more common here than anywhere else in the product industry, followed by cloud and DevOps engineers who run the GPU clusters and model platforms. This is a small Industry Vertical for hiring, made up of many young companies, and Lexsi Labs, a research company in Mumbai, accounts for a large part of the postings. Its hiring is described separately below, so that it does not distort the picture.

Who gets hired, and why

GenAI engineers. A model is only useful once something connects it to people and to other software. GenAI engineers build that connection. They build gateways that send each request to the right model, frameworks for agents that use tools and memory, retrieval that hands a model the right passages from a company's documents, voice agents, and the test harnesses that score a model's answers. Across the young model and platform companies, like Sarvam and TrueFoundry, they are the largest group by far. The postings require Python above all, often with FastAPI, LangChain, Redis and TypeScript, on all three major clouds.

Cloud and DevOps engineers. Models are trained and served on clusters of GPUs, and someone has to build and run those clusters, the Kubernetes platforms on top of them and the endpoints that serve the models. At the platform companies, like Anyscale and H2O.ai, this is a modest group working in Python and Go with Kubernetes, Prometheus and Grafana. Most of the cloud and DevOps postings in this Industry Vertical, however, come from Lexsi Labs, which is building its own infrastructure and hires for configuration and infrastructure-as-code tools such as Ansible, Chef, Puppet, Terraform, Pulumi and CloudFormation.

ML model engineers. These are the engineers who pretrain and fine-tune language and speech models, run distributed training and make models smaller and faster. They work in Python with PyTorch, TensorFlow and JAX. They are a small group, found mainly at Anyscale, a training-platform company, with one opening each at Lexsi Labs and G42, which is a reminder that most foundation models are still trained elsewhere.

Also present in small numbers. Python backend engineers build model and agent APIs. C/C++ systems engineers work at the infrastructure layer. Test engineers test models, labelling tools and agents, and a few Java engineers work at SoundHound AI, the voice-assistant company.

Rare here. Java, frontend, .NET, mobile, data engineering and the enterprise platforms barely appear.

The skills that set this Industry Vertical apart

Python everywhere

Nearly every posting here requires Python, whatever the role.

The tools for building with models

LangChain, model gateways, retrieval and evaluation, in the GenAI postings.

The tools for training models

PyTorch, TensorFlow and JAX, in the ML model engineering postings.

Infrastructure as code for GPU clusters

Kubernetes, Terraform, Pulumi, CloudFormation and Ansible, in the cloud and DevOps postings.

What is changing

These readings follow the record quarter by quarter, from October to December 2025, the third quarter of the financial year 2025-26, to July to September 2026, the second quarter of 2026-27. A longer record will show whether they last.

Hiring held up

The numbers are small, but they held up far better than across product companies when openings fell between July and September 2026.

Persistently open roles

Far fewer jobs here are persistently open than across product companies. Most are filled at the first posting, and Lexsi Labs almost never posts the same job twice. Where a job is posted again, it is most often a GenAI or cloud role at one of the young platform companies.

Where in the Industry Vertical the work is

Market Segment 9.2

MLOps and LLMOps platforms hires most of it. The data and the judge, which covers data labelling, evaluation and alignment research, is its largest sub-segment, because Lexsi Labs sits there. Training and serving and The gateway, the agent and the retrieval hire less each, spread across many young companies, like Anyscale, TrueFoundry and Lyzr AI.

Market Segment 9.1

Foundation model companies hires little. GenAI and ML engineers work at the speech, voice and Indian-language model companies, like Sarvam and Cartesia, alongside the Java engineers at SoundHound AI.

hosted-ai and Wand AI, listed here, have their main homes in Market Segment 5.1 and Market Segment 10.5, and their hiring is described there.

Without Lexsi Labs, Bengaluru holds nearly all the work. With it, Mumbai holds a large share.

The door

The door here is wider than at most product companies. Entry and junior openings are a larger share of hiring, mostly in GenAI work at the young model and platform companies. Lexsi Labs hires mainly experienced engineers, and the rest of the Industry Vertical is more open still.

In one line

If you build applications with language models, the AI-model companies are hiring in India and taking juniors more readily than most. Training foundation models from scratch remains a much smaller door here.

Terms used on this page

Pretraining

is teaching a model from a very large body of data, and fine-tuning is adapting a trained model to a narrower task.

A model endpoint

is the address an application calls to get answers from a running model.

PyTorch, TensorFlow and JAX

are the main frameworks in which AI models are built and trained.

Back to the story

This page reads the openings. The story of the companies is on AI labs and model platforms.