SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Snowflake is seeking a Staff/Principal AI Engineer to join the Cortex Code team, which is building the next generation of coding agents for data work. In this role, you will architect agentic systems at enterprise scale, defining how agents behave reliably, repeatably, and auditably in production environments.
You will own major pillars of the quality stack, tuning agent behavior for advanced coding tasks and designing infrastructure that supports large-scale experimentation. This includes building pipelines and tooling for error mining, prompt/tool/workflow iteration, and establishing clear before/after signals for improvements. You'll lead deep analysis on quality regressions, cluster failure modes, and translate findings into prioritized roadmaps for engineering and modeling teams.
Cross-functional leadership is central to the role. You'll align product, infrastructure, and applied AI on quality standards for critical customer workflows, mentor engineers, and uplevel evaluation practices across the team. You'll ensure quality systems are dependable in practice through reproducible runs, stable datasets, versioning, and operational clarity.
Required qualifications include 10+ years shipping AI/ML-backed software in production with Staff-level ownership of technical direction, cross-team delivery, and mentoring. You need a strong track record building eval harnesses, measurement systems, and experimentation loops for LLM/agent systems—not just one-off benchmarks. Proficiency in Python, TypeScript, and/or Go is essential, along with exceptional communication skills for cross-team influence.
Experience with data engineering pipelines (dbt, Airflow), data modeling, retrieval systems, and semantic layers is a plus. Background in agentic coding tools, LLM observability, safety/guardrails, or quality systems used as release gates is highly valued. You should thrive in high-intensity environments with short feedback loops, care deeply about metrics and reproducibility, and be a power user of modern coding agents.
