About The Role

Builds and deploys AI agents that execute regulated and semi-regulated processes on AWS Bedrock, and owns the document-processing pipelines that feed structured data into those agent workflows — replacing manual steps without introducing RPA-style brittle automation.

Key Responsibilities

Agent Development: Build agents on Bedrock (Amazon Nova and other foundation models) for document generation, data validation, and case intake use cases.
Implement Fargate task logic following established service patterns (e.g., mail-triggered tasks, chatbot-triggered tasks).
Integrate with Microsoft Graph API for Teams-based delivery (message consumption, file attachment handling, callback delivery of outputs).
Write and maintain prompt/agent configurations with appropriate guardrails for regulated activity (e.g., escalation to a human reviewer for non-standard cases).
Unit- and integration-test agent outputs against defined accuracy/turnaround-time KPIs.
Work within a distributed delivery model — code and configuration must be portable across sites.
Document Processing: Deploy and maintain a document-processing Fargate service (chatbot-triggered document processing, output report delivery via callback).
Integrate third-party extraction tooling into the environment as an early-milestone deliverable ahead of full go-live.
Build extraction pipelines for structured and semi-structured documents across varying quality/formats.
Collaborate with the QA/compliance function to ensure extracted data supports audit trail requirements.
Tune extraction accuracy against baseline KPIs before handover to production agent workflows.

Required Experience

8+ years software engineering experience, with 3+ years building LLM-based applications (prompt engineering, RAG, agent orchestration frameworks).
Working knowledge of AWS Bedrock APIs, Python, and event-driven service patterns (SQS/SNS consumers).
4+ years’ experience in document processing/OCR/IDP (Intelligent Document Processing) roles, including hands-on work with AWS Textract or equivalent, plus Azure Document Intelligence.
Experience with financial services or regulated documents preferred.
Domain process exposure is a differentiator, not a requirement — training will be provided via SOP handover.