Role overview :

We are hiring AI Developers (Python, Django, React) to join a client's AI productivity team. The team builds AI agents for business users on AWS, and a major program pilots in November 2026.

What you'll work on :

You will join one of three workstreams, all on the same Python, Django and React stack on AWS.

- AI agents and MCP layer : Build MCP servers and agent back ends with AWS Bedrock and Python.

- Use LLMs (Claude) and Azure Document Intelligence to read, parse and classify emails, POs and invoices.

- AI agent observability : Instrument agents with OpenTelemetry (OTEL). Build tracing, metrics and dashboards that show agent accuracy, cost and performance.

- Feedback and learning : Build the loop where business users rate agent output. Use that feedback, with LLM guidance, to improve prompts. No model fine-tuning is involved.

Across all three, you will :

- Build React front ends and Python/Django APIs end to end.

- Keep agent output accurate and consistent, and watch token cost.

- Ship fast : a new screen requested in the morning should often be ready by end of day.

Required skills :

- Strong Python and Django, including REST APIs.

- React for production front ends.

- Hands-on AWS, especially Bedrock.

- Built at least one real LLM application : prompting, tool calling or agents, and parsing structured data from documents.

- Working knowledge of MCP (Model Context Protocol) or similar agent-tool integration.

- Daily use of AI coding assistants (Claude Code, Copilot, Cursor or similar).

- Clear spoken and written English.

Nice to have :

- OpenTelemetry or other observability tools for AI or distributed systems.

- Azure Document Intelligence or other OCR and document AI services.

- LLM evaluation, prompt versioning or human-feedback loops.

- Awareness of token cost and ways to reduce it (caching, batching, model choice).

- Past work on enterprise email, PO or invoice processing.

What we look for :

Attitude and speed matter as much as coding skill. The right person :

- Works from verbal requirements. Meeting discussions and transcripts often replace formal design documents and turn into stories.

- Asks the right questions early instead of waiting for perfect specs.

- Delivers in hours or days, not weeks.

- Owns the outcome, including accuracy and consistency of AI output.

- Adapts quickly as tools and requirements change.