Backend Developer Intern — Agentic AI Infrastructure

Technology Stack: Python, FastAPI, Async Python, LLMs, Agent Orchestration, Databases
About the Role
We are looking for a Backend Developer Intern to help build the infrastructure behind our production Agentic AI systems.
This is not a sandbox or tutorial-based internship. You will work on real systems that sit underneath AI agents — APIs, orchestration, state management, tool execution, retries, streaming, observability, and failure recovery.
Building backends for AI agents presents different challenges from traditional CRUD applications. LLM calls can be slow, nondeterministic, and expensive. The engaging engineering work is often everything surrounding the model: handling timeouts without duplicating side effects, streaming partial responses correctly, tracking state across multi-step agent runs, recovering from failures, and designing APIs that remain reliable when an external model takes 20 seconds instead of 200 milliseconds.
We are looking for someone who enjoys understanding why a system works, not simply getting it to run.




What You'll Do
• Design and build backend services using Python and FastAPI
• Build APIs that power AI agents and LLM-based applications
• Develop asynchronous workflows for long-running LLM and external API calls
• Build orchestration around agent execution, including retries, timeouts, idempotency, and state management
• Manage state across multi-step agent and tool-calling workflows
• Implement streaming responses for long-running AI operations
• Work with relational and/or NoSQL databases for application data, agent state, and execution history
• Design schemas, queries, migrations, and indexes for real workloads
• Implement rate limiting, particularly around expensive LLM/API operations
• Build logging and observability for agent runs and failures
• Write automated tests covering both successful and failure scenarios
• Debug slow queries, race conditions, flaky integrations, an

📌 Back End Developer (India)
🏢 Raysoft.ai
📍 India