Agentic AI Developer

Role & responsibilities

• Design & implement agentic workflows

  using frameworks like LangGraph, CrewAI(or equivalent), including tool orchestration, planning/replanning, and error recovery.

• Add guardrails & safety

  : Prompt hardening, output validation (JSON schema/Pydantic), policy filters, jailbreak/indirect prompt injection defenses.

• Ground responses with context

  : build RAG pipelines clinical notes, guidelines, and knowledge bases; handle retrieval quality (chunking, reranking), citation & traceability.

• Engineer loops

  : review/approval queues, escalation paths, fallbacks, and exception handling

• Manage agent state & memory

  : conversation/state graphs, checkpoints, and session continuity using Redis/Postgres/object stores; implement short- and long-term memory with expiry and privacy controls.

• Observability & evaluation

  : add traces/metrics, run evals for faithfulness, relevance, and safety (e.g., Ragas), plus task success and Tool Utilization Efficacy.

• Productionize

  : package services (Docker), deploy to cloud (AWS/Azure/GCP), manage secrets, implement role-based access controls, audit trails, and cost/latency budgets.

Preferred candidate profile

• 2 to 3 years software experience (Python preferred) building back-end services or applied AI features; comfort with async I/O, APIs, and testing.
• Hands-on with at least one agent framework (LangGraph, CrewAI) and function/tool calling patterns.
• Practical guardrails experience output schema validation, content/safety filters, and prompt/policy enforcement.
• Context/RAG: vector stores, embedding pipelines, rerankers, and citation-aware generation.
• State management: designing state machines/graphs, persistence (Redis/Postgres/S3), checkpointing, idempotency, and replay.
• Solid CI/CD habits, containerization (Docker), and cloud deployment; observability (logs, metrics, traces).