## About the Role
We are hiring an AgentCore Developer to build, deploy, and operate production AI agents on Amazon Bedrock AgentCore. You will own agent workloads end-to-end — from Runtime and Gateway configuration to Memory, Identity, and Observability — for client engagements that typically begin as a Proof Sprint and graduate into embedded retainers. This is a hands-on builder role for someone who has moved past prototypes and wants to run agents that serve real users under real SLAs.
## Responsibilities
- Design and deploy agents on AgentCore Runtime (serverless microVM sessions, isolated execution, auto-scaling).
- Build tool layers using AgentCore Gateway — wrapping Lambda functions, OpenAPI specs, and internal APIs into MCP-compatible tools.
- Implement AgentCore Memory for persistent, personalized agent context across sessions.
- Configure AgentCore Identity for secure auth to AWS services and third-party systems (Okta, Entra, Cognito, Slack, Zoom).
- Instrument agents with AgentCore Observability (OpenTelemetry traces, dashboards, quality metrics) and integrate with our Magpie governance layer.
- Use the AgentCore CLI and managed agent harness for rapid iteration; drop into Strands-based code when custom orchestration is required.
- Build multi-agent systems using LangGraph, CrewAI, LlamaIndex, or Strands Agents — framework choice driven by the problem, not ideology.
- Ship Code Interpreter and Browser tool integrations for agents that execute code and drive web workflows.
- Own the full lifecycle: prototype → evals → deployment → monitoring → iteration. No throwing work over a wall.
- Work directly with US clients during overlap hours — discovery, scoping, demos, and delivery.
## Requirements
- 3+ years building production backend systems in Python (FastAPI or Django preferred).
- Hands-on AgentCore experience — Runtime, Gateway, and at least one of Memory/Identity/Observability in a deployed project.
- Strong working knowledge of at least one agent framework: Strands, LangGraph, CrewAI, or LlamaIndex.
- Deep familiarity with MCP (Model Context Protocol) — building servers, exposing tools, handling auth flows.
- AWS fluency: Bedrock, Lambda, ECR, IAM, CloudWatch, CDK or Terraform.
- Practical experience with LLM evals — building golden sets, measuring regression, defining quality gates.
- Understanding of agent failure modes: context window management, tool-call loops, hallucinated tool args, retry semantics.
- Solid grasp of containerization (Docker) and serverless deployment patterns.
## Nice to Have
- Experience with fine-tuned SLMs (Qwen, Llama, Mistral) and self-hosted inference (vLLM, Ollama).
- Prior work on HIPAA, SOC 2, or EU AI Act compliant AI systems.
- Contributions to agent frameworks or MCP ecosystem.
- Exposure to LangSmith, Langfuse, or equivalent observability stacks.
- Comfort with US client communication — discovery calls, written updates, live demos.
## What We Offer
- Direct work on AgentCore deployments for named US enterprise and healthcare clients.
- Ownership of agent systems end-to-end — no ticket-shuffling, no design-by-committee.
- Fast promotion path; senior engineers graduate into FDE lead and tech lead roles within 12–18 months.
- Competitive compensation with performance-linked bonuses tied to client retention.
- Hybrid work from our Pune office with flexible hours for US overlap.
- Learning budget for AWS certifications, conferences (re:Invent, AI Engineer Summit), and books.
## How to Apply
- Send your resume + link to an agent system you built — GitHub repo, writeup, or short Loom walkthrough. We weigh what you've shipped more than where you've
Apply to:
[email protected]