SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Empirical Security is building next-generation cybersecurity vulnerability models using ground-truth telemetry to help organizations identify, prioritize, and remediate vulnerabilities across cloud, application security, and traditional infrastructure. The company takes a unique approach by building customer-specific models and maintaining multiple models in parallel.
In this role, you will own the full machine learning lifecycle for exploit prediction models: problem framing, feature engineering, model training, evaluation, deployment, and customer communication. You'll design and train models against real exploitation telemetry, build evaluation frameworks that account for precision, recall, coverage, efficiency, and calibration decay over time, and solve for extreme class imbalance (where only a fraction of published CVEs are exploited in the wild).
Key responsibilities include engineering features across scanner output, EDR, asset inventory, identity, cloud posture, and exploitation telemetry while identifying data leakage. You'll implement the challenging aspects of a dual-model architecture using partial pooling, hierarchical priors, and cold-start strategies for customers with limited initial telemetry. You'll own model monitoring and drift detection, publish research and methodology write-ups, and partner with engineering and forward-deployed teams to move models from notebooks into production systems.
Required qualifications include several years of applied machine learning or statistics with production models that had real consequences, fluency in Python and SQL with strong version control and reproducible pipeline discipline, and deep expertise in classification under heavy imbalance plus at least one of: survival/time-to-event analysis, Bayesian hierarchical modeling, or causal inference. You should have strong calibration instincts, the ability to explain models to security executives while quantifying uncertainty, and curiosity about attacker behavior.
