engineering
Posted 5 days agoStaff AI Product Builder, Data Engineering
at brightwheel
United StatesHybrid
Responsibilities
- Build evidence-first pipelines that produce structured outputs with provenance and uncertainty handling, and that store artifacts rather than overwriting truth.
- - Build a durable job execution system for agent workflows: retries, explicit budgets, idempotency, and monitoring.
- Define success metrics with them, design workflow delivery surfaces, and iterate based on adoption and impact. - Lead by example in AI-augmented engineering, using AI tools to increase velocity while maintaining architectural rigor.
Requirements
- Who You Are You're a Staff-level full-stack builder operating at the intersection of AI systems and data architecture.
- You're AI-native: you understand how LLMs interpret data, and you design retrieval, evaluation, and observability into systems from the start.
- You will: - Ship "virtual employee" workflows that do real work before humans engage: research, verification, prioritization, deduplication, and prep artifacts that cite evidence and flag unknowns. - Design the data foundations that let AI stitch together longitudinal operational signals across domains (customers, prospects, interactions, transcripts, product, ops, billing, support) into reliable workflows.
- - Create shared abstractions for AI and data systems: tool interfaces, logging, cost tracking, evaluation harnesses, data contracts, SLAs, and reusable workflow components that increase trust in both data and AI outputs.
- experience with clear ownership of production systems from design doc through launch and iteration. - A track record of shipping AI-powered workflows to production with measurable impact, including hands-on
- experience with LLM tool use, retrieval patterns, evaluation, and monitoring. -
- Experience operating AI systems in production: evaluation harnesses, rollout strategies, and monitoring that ties system health to output quality. -
- Experience designing data platforms for operational use cases: canonical models, identity resolution and deduplication, and governance patterns that support safe downstream consumption. -