Sarvam AI jobs, judged for you, free

What they do

Sarvam AI builds a full-stack sovereign AI platform in India. It trains its own models for Indian languages and sells APIs for speech, translation, and documents. It also sells agents, on-device models, and consumer apps. Founded in August 2023, it is based in Bengaluru and serves companies and governments.

What they value

• High talent density.
• High ownership from the first day.
• An AI-first way of working.
• A fast pace.
• Own outcomes instead of matching every line of the job spec.
• Systems that run in production, not prototypes.

Where they hire

Every role is full-time and on site in India. Most roles are in Bengaluru. The rest are in Delhi, mostly for Chanakya, a unit that serves the armed forces, intelligence agencies, and ministries. No roles are remote.

Teams and roles

Engineering has the most roles, including staff roles on the API platform, product infrastructure, and apps.

• API platform: tens of millions of calls a day over HTTP and WebSockets, moving from Python and FastAPI to Go, with RFCs before code.
• Document intelligence: Go, Python, Temporal, Kubernetes, and PostgreSQL, with React and TypeScript review tools for long PDFs.
• Apps: Kotlin Multiplatform on five platforms, for low-end Android phones on poor networks.
• Inference: source-level work in vLLM or SGLang, GPU kernel performance, and HPC site reliability.
• Agents: versioned code with evals and regression suites.
• Also Chanakya backends for air-gapped settings, foundation model research, data, security, and DevOps.

Product managers talk to users, write specs, ship every week, and use SQL and PostHog or Mixpanel. The PM role is not for people from consulting or program management. The product designer needs a portfolio, and Claude Code and Cursor are a strong plus.

What they look for

Sarvam wants people who shipped systems and kept them running in production. It values measured results over activity, hands-on use of AI tools, and interest in India's languages and conditions. Product managers must understand model latency, cost, failure modes, and evals. The site does not describe the interview steps.