The company is looking to convert research Proofs of Concept (POCs) into production-grade services with Service Level Agreements (SLAs), on-call readiness, and guardrails, specifically focusing on building agentic workflows and robust Retrieval-Augmented Generation (RAG) pipelines.
Requirements
- 5+ years building production web apps/services; 1-2+ years with LLMs/RAG/agents.
- Strong Python (3.10+), FastAPI, async patterns; TypeScript with React/Next.js.
- Solid RAG fundamentals (indexing, embeddings, hybrid search, re-ranking, evals) and Agent patterns (tooling, retries, timeouts, circuit breakers).
- Experience with vector databases (pgvector/Milvus/Weaviate) and Postgres.
- CI/CD (GitHub Actions), Containers/K8s, IaC (Terraform), and cloud (AWS/Azure/GCP).
- Testing discipline: unit/integration/e2e, load tests, contract tests; strong debugging & profiling.
- Add observability (OpenTelemetry, logs/metrics/traces, cost & token spend) and performance/cost tuning.
Responsibilities
- Convert research POCs into production-grade services with SLAs, on-call readiness, and guardrails.
- Design and ship agentic workflows (tool use, function calling, orchestration graphs, retries, fallbacks).
- Build robust RAG pipelines: ingestion → chunking → indexing → retrieval → re-ranking → grounded generation.
- Implement evaluation & LLMOps: golden sets, regression tests, CI eval gates, telemetry dashboards, A/B and canarying.
- Develop React/Next.js frontends for multi-step agents, streaming outputs, uploads, and auth flows.
- Build and document FastAPI backends (async, OpenAPI, pagination, rate limits, idempotency, background jobs).
- Operate via Kubernetes with Docker images, IaC (Terraform), GitHub Actions, and S3/GCS/Azure Blob storage.
Other
- Proven end-to-end ownership: shipped at least one LLM/Agent system to real users.
- Clear communication, product sense, and trade-off thinking (latency/cost/quality/safety).
- Partner with Product, Sec/Compliance, and Platform to meet privacy, audit, and governance needs.
- Establish coding standards, testing strategy, and incident/rollback playbooks.
