Applied AI Engineer, Sarvam Agents

Sarvam is building India's full-stack sovereign AI platform across research, models, infrastructure and applications, backed by Lightspeed, Peak XV, and Khosla Ventures. Sarvam Agents plug into the tools and workflows of an organization (email, docs, calendars, internal systems, the long tail of SaaS), take multi-step action on a user's behalf, run on a schedule when no one is watching, and get better as the team uses them.

Responsibilities:

• Build agents end-to-end: flow design, prompting, tool definitions, memory, evaluation, deployment
• Own agent reliability and runtime management in production, including state handling, retries, and scheduled execution
• Build connector frameworks and OAuth integration for third-party SaaS APIs
• Scale MCP server infrastructure and optimize the tool layer
• Engineer memory systems (working, long-term, and retrieval-based)
• Implement multi-tenant agent sharing with permissions and audit capabilities
• Build evaluation infrastructure and prompt iteration workflows
• Optimize cost and latency for live traffic

Requirements:

• 3-5 years total experience, with 2+ years strong Python backend development
• LLM API fluency: prompt design, structured outputs, function calling at production quality
• Shipped AI features to real users in production, not just notebooks or prototypes
• RAG patterns: chunking, embeddings, hybrid search, reranking, vector databases
• Experience building and running AI system evaluations
• PostgreSQL/MySQL and Redis expertise
• Observability platforms (Signoz, Datadog)

Named tools: LangGraph, ADK, MCP servers, Honcho (memory), LangSmith and Braintrust (evals), Signoz/Datadog (observability), vector databases.

Location: Bengaluru, Full-Time, On-Site.