Sarvam is at the forefront of building India's sovereign AI platform, focusing on research, models, infrastructure, and applications to make AI impactful for the nation. We collaborate with major enterprises and public institutions, backed by prominent venture capital firms. Our work involves partnering with leading Indian brands to develop cutting-edge AI solutions.

At the core of our conversational AI platform is a robust system enabling customers to build, configure, deploy, and manage voice agents and campaigns at scale. This system currently handles over 50 million conversation minutes monthly and is experiencing rapid growth.

As a Staff Engineer, Product Infrastructure, you will own the end-to-end backend system responsible for agent, campaign, and user management. This includes designing the data model, developing APIs, managing orchestration, and ensuring high reliability and performance.

You will be responsible for workflow orchestration using Temporal for complex, stateful operations. Enhancing observability through logging, metrics, and tracing to enable swift debugging will also be a key focus. Furthermore, you will develop integration testing infrastructure to facilitate rapid, safe deployments and collaborate closely with the frontend team to deliver an exceptional user experience.

We are seeking an experienced Staff Engineer with 5-10 years of experience in building production backend systems, demonstrating deep expertise in Python and FastAPI (or Golang). A proven track record of full-stack ownership, including system design, trade-off analysis, and production operation, is essential.

Hands-on experience with large-scale workflow orchestration, preferably Temporal, is required. Candidates should have experience building and operating high-scale distributed systems, including multi-tenant, low-latency, and resilient platforms. Strong fundamentals in PostgreSQL and Redis, covering schema design, query optimization, and caching strategies, are necessary. Excellent judgment in making build vs. buy, optimize vs. ship, and abstract vs. inline decisions is also expected.