About the team
Sarvam builds Indic-language AI used at population scale. The CPU Infrastructure team runs everything outside the GPU fleet: Kubernetes across Azure, AWS, GCP and Indian sovereign clouds, the network between them, Cloudflare, identity, secrets, observability, cost and the delivery pipeline every product ships through.
The team builds and runs what it builds. Engineers own real systems end to end, and every repeated piece of manual work is a candidate for automation. As an AI company, we are building AI agents into our own operations, to plan and review infrastructure changes, triage alerts and handle routine issues, with engineers setting the guardrails.
About the role
You keep the infrastructure behind Sarvam’s products running. You handle day-to-day operations — deployments, provisioning, access, incidents and cost questions — and you are expected to automate that work away.
We hire for one instinct above everything else: you notice something done by hand every day, ask why it can’t run itself, and build the thing that removes it — a script, a pipeline step, a bot, a small Kubernetes operator, or an AI agent that does the work. Operations is a large part of this role, and it is where you find what is worth automating.
As you automate your own toil away, more of your time goes into platform and automation work.
Products you may work on
Sarvam builds across a broad range of AI products: voice agents, coding agents, coworker agents, dubbing, our API platform, and the real-time voice and WebRTC infrastructure underneath them, among many others. You will work on the infrastructure behind them.
Responsibilities
• Own infrastructure operations across Sarvam’s products: deployments, provisioning requests, access, environment issues, cost questions
• Respond to incidents, write the runbook or postmortem afterwards, and fix the cause
• Automate repeated work: turn manual steps into pipeline jobs, Terraform modules, templates or self-service for product teams
• Build AI-assisted automations: agents that triage alerts, summarise incidents, review routine changes or answer product teams’ infrastructure questions, with a human approving anything risky
• Security basics: keep images and dependencies patched, rotate credentials, keep secrets out of code, fix security findings
• Reliability basics: keep dashboards and alerts accurate, track SLOs and cost, delete noisy alerts
• Take part in on-call
Requirements
• 2–5 years in DevOps, SRE, infrastructure or backend engineering
• Has built infrastructure outside a ticket: a homelab, an open-source contribution, a side project, or an internal tool nobody asked for — and can show the code (GitHub, a blog post or a demo)
• Comfortable on the Linux command line: processes, networking basics, logs, systemd
• Working knowledge of Docker and Kubernetes; has deployed and debugged something running on a cluster
• Scripting in Python, Go or Bash
• Basic Git and CI/CD experience
• Has spotted a repeated manual task — a daily review, a recurring fix, a routine request — and built something that removed it: a script, a CI job, a bot or a Kubernetes operator. You can walk us through what you noticed, why it mattered, what you built and what it saved
• Has used LLMs or AI agents to automate real work, not only for writing code
• Fixes causes rather than the same symptom twice; asks “why is this manual?” before doing it again
• Curious, takes ownership, wants to grow into platform engineering
• Nice to have: Terraform, a public cloud (AWS, Azure or GCP), Prometheus / Grafana, exposure to SIP, WebRTC or real-time media, any cloud or Kubernetes certification