Claude Code Backend Developer

Описание

Responsibilities

• Develop the backend of AI functions: chat, streaming responses, retrieval, Deep Research
• Work with LLM provider APIs: parallel calls, limits, prompt caching, cost
• Define contracts between frontend and backend and know where to put checks
• Work with prompts, context, retrieval and Chroma
• Build CLAUDE.md, skills
• Review AI-generated PRs quickly and systematically, not read sequentially
• Find defects beyond green tests
• Analyze bugs that AI does not reach or that it confidently "fixes" incorrectly
• If there is a problem on the frontend, understand React + TypeScript and fix it
• Make technical decisions and be responsible for the result. Every day you work in conjunction with our lead backend/AI engineer and remove everything that prevents them from doing the most complex tasks. We are at the launch stage, and we need an expert, not someone who will figure things out on the go.

Requirements

• At least three years of commercial backend development experience, including LLM services in production
• Python and asynchronous programming: asyncio (gather, TaskGroup, Semaphore, task cancellation, timeouts); difference between async, threads and processes, GIL, when to use what; httpx or aiohttp, connection pooling
• FastAPI: async endpoints, Depends, Pydantic v2; streaming (StreamingResponse, SSE, background tasks); handling client disconnections
• LLM provider APIs (OpenAI, Anthropic, Gemini): streaming, tool use / function calling, structured output; prompt caching, token counting, usage; rate limits (RPM, TPM), 429, retries with backoff; understanding that different providers have different formats, and experience with abstraction over them (LiteLLM or your own)
• Data: PostgreSQL (transactions, indexes, concurrent writes); Redis (cache, locks, queues, pub/sub)
• Frontend: understanding React + TypeScript to the extent that you can fix it yourself in case of a major problem
• English B1+: read and write technical documentation

Nice to have

• RAG: embeddings, vector databases (Chroma, pgvector, Qdrant), chunking
• Queues: Celery, arq or Dramatiq for long tasks like Deep Research
• Observability: LLM call logging, tracing (Langfuse, OpenTelemetry), cost accounting
• Infrastructure: Docker, nginx (settings for SSE), basic CI
• Evals: how to check that answers have not worsened after changing prompts
• Tests: pytest, LLM call mocks

We offer

• Full-time, remote work.
• Salary will be discussed at the interview.
• Option program included.

Hiring process

• No resume required. No cover letter required. Go to our HR bot: answer questions verbally, concisely and to the point. Do not use Claude in your answers: all your answers are given with a microphone pressed to your ear, preferably without pauses or freezes.
• Wait for you after successfully completing a short dialogue with our HR bot.