Design, build, and maintain the backend services, APIs, data-platform automation, and AI agents behind our data products. This is a backend-heavy role: distributed Python services, streaming LLM/agent runtimes on AWS, RAG pipelines, and the dbt/Airflow automation that powers our data products. Responsibilities • • Design, build, and maintain backend services and APIs in Python — RESTful and streaming (SSE) endpoints, agent runtimes on AWS Bedrock AgentCore / ECS Fargate, and event-driven Lambda handlers. • Architect service boundaries and data flows: define contracts between services, model persistence, and manage state, caching, and asynchronous/background processing. • Build RAG pipelines and tool-calling AI agents over our data products: retrieval, orchestration, grounding, and evaluation. • Design data access and storage layers: schema/data modeling, query performance, connection/session management, and integration with warehouses (Snowflake) and key-value stores (DynamoDB). • Implement auth and identity: OAuth/OIDC flows (per-user 3LO, token vaulting, session binding), least-privilege IAM, secrets management. • Build and maintain data-platform automation: dbt models, MWAA/Airflow orchestration, and tooling for discoverable, governed, consumable data products. • Own service reliability and delivery: Terraform, GitHub Actions CI/CD, container builds, structured logging, metrics/tracing, alerting, and cost controls. • Set technical direction: system and API design, code review, and mentoring. Required skills Backend engineering • • 5+ years designing, building, and maintaining production backend services at scale. 10+ years if Lead level engineering candidate. • Expert-level Python for server-side development; solid grasp of at least one web/async framework (e.g. aiohttp, FastAPI, Flask) and the WSGI/ASGI model. • Service and API design: REST (and/or gRPC), request/response and streaming patterns, pagination, versioning, idempotency, and backward-compatible contracts. • Data layer: SQL and data modeling, query optimization and indexing, transactions, connection pooling; experience with relational, warehouse (Snowflake), and NoSQL/key-value (DynamoDB) stores. • Server-side patterns: caching strategies, background jobs/workers, queues and event-driven processing, rate limiting, retries/backoff, and timeouts. • Performance & reliability: profiling, load handling, latency/throughput trade-offs, graceful degradation, and designing for failure. • Observability: structured logging, metrics, distributed tracing, and debugging live production issues. Software engineering fundamentals • • Object-oriented programming (required): encapsulation, abstraction, inheritance, composition, polymorphism; SOLID principles; design patterns applied pragmatically; strong domain modeling. • Solid data structures & algorithms; ability to reason about time/space complexity. • Concurrency and async programming (async/await, threading, event loops) and their failure modes. • Testing (unit, integration, end-to-end) and testable design; Git and PR-based workflows; disciplined code review. Cloud & infrastructure • • Production AWS: ECS/containers, Lambda, IAM, API Gateway, DynamoDB. • Infrastructure as code with Terraform; CI/CD (GitHub Actions or equivalent) and container builds. Distributed systems • • Building services that are horizontally scalable, resilient, and loosely coupled; handling consistency, retries, idempotency, and partial failure. Security • • OAuth/OIDC, authn/authz, token handling, least-privilege access, multi-tenant isolation, secrets management. AI / LLM engineering • • Building LLM applications in production (not research). • RAG: chunking, embeddings, vector search, hybrid search, reranking, grounding/citations, context-window management, retrieval evaluation. • Agents: prompt engineering, tool use/function calling, structured outputs, agent orchestration (single- and multi-step), prompt caching. • Integration with model providers — Anthropic/Claude, AWS Bedrock/AgentCore, Snowflake Cortex — and response streaming. • AI quality & ops: eval harnesses, guardrails, tracing/observability, token/latency/cost optimization. • AI security: prompt injection, data exfiltration, PII handling, per-user identity/RBAC enforcement. Nice to have • • dbt, Airflow/MWAA, Snowflake. • MCP (Model Context Protocol) and/or RAG • Slack platform (Bolt, Socket Mode, Block Kit) or other real-time/conversational backends. • Fine-tuning/adaptation, semantic caching, or model routing/fallback. • Experience adding AI capabilities to existing production systems.