Mid+ AI / Backend Engineer

About the project

For a fast-growing startup building a B2B AI platform on autonomous agents, we are looking for a Mid+ AI / Backend Engineer.

The client's platform takes tedious analytics off enterprise teams through a flexible architecture, agentic solutions and advanced workflows. You will join the engineering team with real influence over the architecture of systems that go straight to production: you build and scale both the backend infrastructure and the AI systems.

We are looking for someone who is not a senior engineer yet but is close to it. The client wants to work with someone independent: you get a well-described task and you deliver it without being walked through it.

Day to day

• Build and grow scalable backend services (Python, FastAPI) that power the AI platform.
• Build and optimise RAG pipelines, AI agent systems and LLM orchestration in production.
• Help build the cloud infrastructure (AWS / Azure), including multi-tenant isolation and serverless functions.
• Build a flexible integration ecosystem driven by APIs and third-party connectors.
• Review code, keep testing standards up and mentor junior engineers.
• System security: JWT, OAuth2, RBAC, protection against prompt injection.

Requirements

• 3-5 years of commercial backend experience in Python.
• Strong FastAPI (async, Pydantic, dependency injection).
• PostgreSQL: schema design, query optimisation, migrations (Alembic).
• AI systems (LLM, RAG): tool calling, function calling, multi-step reasoning, MCP (Model Context Protocol).
• Cloud: AWS (Lambda, S3, ECS, API Gateway) and/or Azure.
• Docker, Kubernetes, CI/CD (GitHub Actions, GitLab CI), IaC (Terraform / AWS CDK).
• Production experience integrating LLM APIs (OpenAI, Anthropic Claude, Google Gemini).
• Strong written and spoken English.

Nice to have

• TypeScript with an advanced type system (generics, utility types, conditional types).
• Celery and asynchronous task processing (Redis / RabbitMQ).
• GraphRAG and graph databases (Neo4j).
• Serving LLMs on-premise (vLLM, TGI, TensorRT-LLM), fine-tuning (LoRA / QLoRA).
• Pulumi, gRPC or GraphQL.

The process

• Screening call with CRODU (about 20 minutes).
• Technical conversation with the client's team.
• Decision and start.