Abnormal
Machine Learning Engineer II
About this role
Abnormal AI seeks a Machine Learning Engineer II to join the Attack Detection team, building high-recall email threat detection systems that process hundreds of millions of messages with millisecond latency. You'll design and optimize detection pipelines combining rules, models, and feature engineering while continuously adapting to evolving attack tactics.
What you'll do
- Design and implement detection systems combining rules, ML models, and feature engineering with guidance from senior engineers
- Engineer features and model approaches to improve detection efficacy for email attacks
- Train and evaluate models on specialized datasets while monitoring false negative and false positive rates
- Analyze failure cases to identify capability gaps and recommend feature or rule improvements
- Build and maintain data pipelines for model training, evaluation, and deployment
- Collaborate with infrastructure and systems engineers to productionize detection signals
What they're looking for
- Machine Learning (3+ years in NLP, text understanding, entity recognition, or similar domains)
- Python and ML frameworks (NumPy, scikit-learn, PyTorch, TensorFlow)
- SQL, Pandas, and Spark for data pipeline and analytics work
- Production ML pipeline development and model evaluation
- Software engineering best practices and debugging
- Email security or threat detection domain knowledge
- Big data and statistical analysis
- Algorithm design and optimization
Opens the official application on the employer’s site. No login required.
Abnormal
Abnormal builds AI-powered email security solutions that protect against threats through narrative-driven threat understanding, SMTP relay infrastructure, and advanced filtering capabilities. The company is hiring software engineers, infrastructure engineers, sales engineers, and technical support specialists to expand its platform and support enterprise customers.
View all jobs at AbnormalLikely interview questions
- Describe your experience deploying machine learning models to production—what were the biggest challenges you faced?
- Walk us through how you've debugged a production ML system where model performance degraded unexpectedly.