AI Engineer

Role Summary

We''re hiring an AI Engineer to build a production-grade AI agent that automates business-analyst workflows inside an enterprise SaaS environment. The agent ingests meeting transcripts and supporting documents, reasons across existing knowledge sources (Jira tickets, Confluence pages, Salesforce metadata), and generates structured Business Requirement Documents and Jira user stories with a human-in-the-loop review at each stage.

You will build the agent end-to-end: connectors, retrieval, prompts, evaluation, and deployment. This is a hands-on coding role, not a prompt-tweaking one. The first deliverable is a working agent in roughly four weeks; the engagement is expected to extend into additional agent use cases beyond the initial scope.

What You''ll Do

• ·       Build an LLM-powered agent on Anthropic Claude that reads inputs from a Jira ticket, processes them, and writes outputs back to the same ticket.
• ·       Implement retrieval over Jira tickets, Confluence pages, Salesforce metadata, and other GTM knowledge sources — start simple (search + fetch + context assembly), layer in RAG/vector retrieval as the use case demands.
• ·       Design the agent''s reasoning loop: when to ask clarifying questions, when to commit to a BRD draft, how to detect scope creep when new inputs arrive.
• ·       Build the BRD-to-story decomposition: epics, user stories, acceptance criteria — output structured to a BSA-provided template.
• ·       Write evaluations from day one. Golden-dataset coverage for BRD quality, faithfulness checks against source documents, regression tests when prompts change.
• ·       Implement audit logging — every input, retrieval, model call, and output must be traceable.
• ·       Deploy and operate the agent. CI/CD, monitoring, basic cost tracking.
• ·       Iterate based on BSA feedback in the Jira flow. The real-world use case is messy; the agent has to gracefully handle re-uploads, new documents, and scope changes.

Required Experience

• ·       5+ years of software engineering. Production code you can talk through.
• ·       Strong Python. Comfortable in TypeScript or Node for tooling/UI work.
• ·       Git, CI/CD, Docker. Can read and make small changes to Terraform.
• ·       Has shipped at least one service to production with logging, monitoring, and tests.
• ·       Hands-on with Anthropic Claude (preferred). Exposure to GPT-4o or Gemini is fine; the project standardizes on Claude.
• ·       Has built an agent or agentic workflow using a real framework (LangGraph, Google ADK, CrewAI, or hand-rolled) — can explain trade-offs between single-agent and multi-agent designs.
• ·       Production RAG experience: chunking strategies, embedding selection, hybrid retrieval, reranking. Knows when RAG is the wrong answer.
• ·       Vector DB experience with at least one of: pgvector, Pinecone, Weaviate, Vertex AI Vector Search.
• ·       Comfortable integrating with enterprise APIs: REST, GraphQL, OAuth.
• ·       Has worked with at least one of: Jira API, Confluence API, Salesforce API. Direct experience with any of these is a strong plus.
• ·       Working knowledge of at least one major cloud (AWS, Google Cloud Platform, or Azure).
• ·       Container/Kubernetes basics (Docker, plus AKS / GKE / ECS at a working level)

Nice-to-Have

• ·       Direct experience with Atlassian APIs (Jira, Confluence) or Salesforce Apex/REST APIs.
• ·       Built or contributed to internal AI tooling — eval harnesses, prompt registries, observability for LLM apps.
• ·       Workato or similar iPaaS / workflow automation exposure.
• ·       Domain familiarity with Quote-to-Cash, CPQ, CLM, or other GTM/RevOps systems.
• ·       Experience working in regulated environments (SOX, SOC 2, FedRAMP, HIPAA, or equivalent).
• ·       Open-source contributions to LLM/agent frameworks.
• ·       Has demoed their own work to a customer or executive audience.
•