- Build and ship automations and agentic workflows that call our internal REST and GraphQL APIs to read data and take production actions.
- Deliver document automation (OCR and extraction) using managed services and LLM vision. This includes confidence thresholds, deterministic validation against TMS data, and routing low-confidence results to human review.
- Determine and design the right approach for each workflow: LLM, deterministic software, RPA, or a purchased tool, with clean handoffs between them.
- Run build vs. buy evaluations, including POCs and bake-offs covering total cost of ownership, integration effort, and operating burden. Then implement the decision, whether that means building or integrating and rolling out what we buy.
- Partner closely with our RPA team (UiPath). Identify existing bots that should be augmented or replaced with AI, and design workflows where AI and RPA hand off cleanly.
- Coordinate with Engineering and Platform when automation needs platform changes. Scope the ask clearly, then deliver your side.
- Build the evaluation infrastructure for every AI system we ship: labeled datasets, LLM-as-judge pipelines, regression testing, and release gates. Product defines the quality criteria; you make them executable.
- Make AI systems dependable through validation, fallbacks, idempotency, permissions, audit trails, monitoring, and capture of reviewer corrections.
- Inventory and rationalize our existing low-code automations (Zapier): decide which to keep, harden, consolidate, or replace. Check what our existing stack already covers before adding anything new, and avoid depending on tools we plan to retire.
- Prototype quickly against real workflows and data, and take successful work all the way to production.
Scope
This role consumes platform capabilities and does not own core platform, authentication, or infrastructure. Agent infrastructure needs, such as tool authorization and observability, are evaluated for buy or partner options first, in partnership with Engineering and DevOps. Initial document AI work uses managed services; model hosting and training are in scope only if evaluation shows a clear need.
All production work follows Engineering's standards: code in our shared repositories, Engineering code review and architecture review, standard deployment and monitoring, and Engineering sign-off on production readiness. Every automation is documented with an owner, its API dependencies, and a runbook, so any engineer can operate and inherit what you build.