Tech Lead Python / LLM Agents

Senior Tech Lead Python / LLM Agent role, remote or hybrid, full-time, 6+ years experience, leading platform core and agent cycle.

Core platform development in Python: sessions, agent steps, customer policies, accounting; agent loop under step budget. Migrating users from pilot architecture to target without downtime. Reducing step cost: input composition, prefix cache, tool limits, model routing, per-user and per-step accounting. Quality assessment: reference from real traces, comparison of Russian models, regression on each prompt, tool, model. Company data via MCP with user permissions: 1C, KHD, SED, HR, document search with ACL. Reliable background procedures and writes to external systems: schedules, human confirmation, exactly-once writes. Sandbox for agent code: per session, no secrets and outgoing connections. Team: decomposition, review, work with AI agents for code, hiring. Aligning requirements and security with the customer.

Backend 7+ years, with at least 2 years as tech lead/principal engineer responsible for production. Python in production: asyncio, FastAPI, Pydantic, typing, tests. Able to read and modify Go. LLM agents in production with real users: tool calling, agent loop, constraints, error handling. Understanding context and cost: call input, prefix cache, history compression, tool limits, calculate step price based on usage. Eval: reference sets, LLM judge and its validation, regression, tracing (Langfuse, OpenTelemetry). Hybrid search, reranking, chunking, document access rights. MCP, OAuth 2.0/OIDC, token exchange — agent on behalf of user. Durable execution (Temporal, Restate or similar), idempotency, retries. PostgreSQL, Redis, S3, Docker, Kubernetes, GitLab CI. Open models via OpenAI-compatible APIs, LiteLLM or similar. Higher technical education.

Full-time employment, schedule 5/2, remote or hybrid work format.