Sr. Software Engineer, Core AI

About the Role

We are building a fleet of specialized AI services that improve high-value human workflows across the company. As a Senior Software Engineer, Core AI, you will design, build, evaluate, and operate production-grade AI services and agents. You will own work end to end—from data and context design through tool use, application logic, evaluation, and production monitoring.

What You’ll Do:

• Build specialized AI services: design and implement task-specific services and agents, including data serialization, retrieval, memory where appropriate, tool integrations, and bounded multi-step reasoning.
• Define and evaluate quality: work with partners to make expected behavior explicit, write and maintain evals against representative examples, and use results to measure performance against expectations.
• Improve quality systematically: use evaluation findings, failure analysis, and production feedback to improve context and data, service logic, tools, and model behavior.
• Build for production: integrate tracing, observability, versioning, monitoring, reliability practices, and success metrics into the services you own.
• Run controlled experiments: compare models, data, and implementation approaches using repeatable benchmarks so improvements are attributable and reproducible.
• Partner on architecture: work with AI Architecture, AI Ops, and Platform teams to make services secure, reliable, cost-aware, and deployable at scale.
• Ship iteratively: move quickly from user problem to prototype to production, learning from real workflow outcomes.
  

What We’re Looking For:

• 5+ years of strong software engineering experience delivering production services or products, ideally at senior or staff level.
• Hands-on experience building AI/LLM-powered applications or agents with real users or operational use.
• Experience with evaluation methods for AI behavior: defining expected outcomes, building test sets or evals, tracking regressions, and acting on results.
• A data-first mindset: able to shape retrieval, context, and serialization to give a model the information needed for the task.
• Good judgment about tool use, autonomy, human review, failure handling, privacy, reliability, latency, and cost.
• A product-minded approach focused on solving specific workflow problems and measuring whether the solution works.
• Comfort learning tools and frameworks; experience with agent runtimes, model APIs, MCP, tracing, or evaluation tools is helpful.