Senior Software Engineer—AI Product & Platform Engineering

AAPC•Salt Lake City, UT

5h•Remote

About The Position

We build AI-powered products that improve customer workflows (speed, accuracy, confidence). As we deliver outcomes, we extract the repeatable patterns into an internal AI engineering platform that makes the next outcome faster and safer. This is not a "build it and they will come" platform role; platform adoption is earned through shipped product value.

Requirements

Senior-level TypeScript + React experience in production systems
Experience building React full-stack apps (routing, server rendering, server components/actions, backend-for-frontend patterns)
Async + streaming experience (SSE/streams; cancellation/backpressure awareness)
Comfortable on macOS and Linux; fluent with CLI for local dev, debugging, automation, and as an API
Strong Git discipline (PR workflows, code reviews, conventional commits, clean commits) and ability to raise the bar through review
Strong testing discipline and ability to build testable abstractions
Strong API integration experience; ability to ship, measure, and iterate in ambiguity

Nice To Haves

Experience with the Vercel AI SDK (Core + UI) for streaming experiences, tool calling, and chat UX patterns
LLM integration experience (OpenAI/Anthropic/Bedrock or similar) with tool calling / structured outputs
Experience building internal platform primitives with real adoption (SDKs, shared libraries, paved roads)
Experience with AI evaluation frameworks and regression testing for model outputs
Experience with API design and an intuition for
Observability and/or security depth for AI systems

Responsibilities

Ship end-to-end product increments (spec → build → release → operate)
Use the right tool for the job; our current default is a modern TypeScript + React full-stack (server rendering, RSC-style patterns, streaming), and we stay flexible as we learn
Build AI features with disciplined patterns: tool calling, structured outputs, grounding, and streaming UI
Partner with product, design, and SMEs to define outcomes, validate assumptions, and iterate quickly
Turn lessons from shipped features into reusable "paved road" primitives (eval harnesses, guardrails, shared tool/prompt patterns)
Build evaluation + release loops (tests, golden datasets, regressions, targeted human review; LLM grading where it fits)
Own reliability, performance, and cost; instrument with OpenTelemetry and define SLOs for key flows
Enforce security and privacy by design (safe tool access, authZ, auditability, prompt-injection mitigations, PII handling)