Staff Security Software Engineer - Agentic Security Engineering

Databricks

1d•$152,000 - $261,250•Remote

About The Position

The Agentic Security Engineering team is a horizontal, shared services engineering team that makes every Databricks security function AI-native — building the AI agents, "personas," and shared platform that security teams across the org (Detection & Response, Threat Intelligence, Vulnerability & Product Security, Red Team, GRC, and Continuous Monitoring) build on to operate at machine speed, safely and reliably. We leverage AI to transform how security is done. As Staff Security Software Engineer, you own a major technical workstream designing and shipping production-grade AI agents and the platform capabilities that power them, with reliability, observability, and safe-by-default security as first-class requirements. You set the standards for how agents are built and operated and co-build with partner security teams. Example use cases you'll enable, led by AI threat detection at scale (alert triage and false-positive reduction, fleet-scale anomaly detection), extend to threat hunting, vulnerability management, product security, red team/offensive testing, and GRC/assurance use cases.

Requirements

Demonstrated experience building, shipping, and operating production AI agents that others depend on, with an intentional pivot to AI-native transformation of a function at scale (prototypes and demos don't count).
Expert-level Python for production systems at scale.
Depth across the production-AI dimensions: agent architecture & orchestration; context & memory management; production operations (latency, cost, caching, monitoring, model upgrades, reliability); and AI quality & evaluation (eval frameworks, LLM-as-judge, regression testing).
Hands-on experience with modern agentic frameworks and LLM tooling (e.g., LangChain, Hugging Face, frontier models) and MCP-style tool integration.
Strong technical communicator who aligns partners and drives adoption without direct authority.
7+ years in software or security engineering, with demonstrated technical leadership on a significant workstream.
Track record of shipping high-quality security tooling that other teams depend on in production.

Nice To Haves

Research contributions or deep familiarity with adversarial ML, AI safety, or AI red-teaming methodology.
Experience with MLOps platforms, AI serving infrastructure, or AI platform security at cloud scale.
Familiarity with AI governance standards (NIST AI RMF, ISO/IEC 42001, EU AI Act technical provisions) as they apply to security engineering.
Open-source contributions or publications in AI security, adversarial ML, or security tooling.

Responsibilities

Design, ship, and operate production AI agents/personas that security teams depend on, starting with AI threat detection at scale.
Build and harden the shared platform so any security team can ship agents safely and reliably — i.e., sandboxing, scoped least-privilege identity, observability, and a reusable agent/tool catalog.
Set engineering standards and make reliability first-class — design review bar, latency/cost optimization, monitoring, clean CI-based deploys, and safe model upgrades.
Establish AI quality and evaluation practices: eval frameworks, LLM-as-judge (with bias controls), and regression gates that catch quality drops before production.
Lead the design and development of AI platform capabilities that operate at production scale — behavioral analysis of usage, detection of prompt injection attempts, and anomaly detection on agentic workflow behavior.
Define the methodology for AI security assessment: how Databricks systematically evaluates new AI capabilities against a comprehensive threat model before deployment and monitors them continuously after.
Drive technical strategy for AI red-teaming tooling: automated adversarial testing platforms that simulate how real attackers attempt to abuse Databricks' AI systems.
Co-build with and enable partner teams across the security org, and mentor mid-level engineers.
Lead design reviews, define team engineering practices, and drive continuous improvement in the quality and reliability of AI security tooling.