What you’ll do
- Build and extend FastAPI services that back the agent platform: organizations, projects, billing, telephony, and the admin console
- Work on the LangGraph agent-execution engine — tool calling, escalation, conversation state, and the config-driven graph builder
- Design PostgreSQL schemas and write Alembic migrations that hold up under multi-tenant load
- Integrate voice carriers (Twilio, Plivo) and LLM providers (OpenAI, Anthropic, Gemini) behind consistent internal interfaces
- Wire up RAG retrieval over Qdrant and analytical queries against BigQuery/Snowflake for customer-facing analytics
- Add timeouts, retries, and observability (Prometheus, OpenTelemetry) so failures in one dependency don't take down a call
- Ship changes to a Kubernetes-deployed service with real on-call implications — you'll own what you build in production
What we’re looking for
- 3+ years building production backend services in Python
- Solid experience with FastAPI or a comparable async web framework, plus SQLAlchemy/Alembic or similar ORM + migration tooling
- Comfortable designing relational schemas and writing migrations you'd trust against a live, multi-tenant database
- Working knowledge of Docker and Kubernetes-based deployment
- Experience integrating third-party/webhook-based APIs (payment, telephony, or similar) and handling their failure modes
- You reason about concurrency, timeouts, and retries by default, not as an afterthought
Nice to have
- Experience with LLM-based agent frameworks (LangGraph, LangChain) or building tool-calling agents
- Familiarity with vector databases (Qdrant, pgvector) and RAG pipelines
- Experience with real-time/streaming systems — WebSockets, telephony media streams, or similar
- Exposure to Terraform, Vault, or infrastructure-as-code in a cloud environment (Azure/AKS)