Overview As Senior AI Scientist in Trust & Safety, you design and deploy fraud-detection models that operate end-to-end-from data discovery to monitoring in live environments. You develop real-time risk scoring, graph-based anomaly detection, and generative/agentic AI to catch bad actors while reducing customer friction. You work with cross-functional teams to translate model outputs into enforcement actions and scalable improvements. This role offers impact at scale in a mission-driven fintech setting and the chance to shape how Intuit combats fraud with advanced AI.
Compensation / Benefits- competitive compensation
- cash bonus eligibility
- equity rewards
- comprehensive benefits package
Responsibilities- Design, build, and deploy fraud/abuse detection models for real-time and batch use across customer lifecycle
- Own end-to-end model lifecycle: data discovery, ETL, feature engineering, training, productionization, monitoring, and retraining
- Explore advanced detection methods: entity-level anomaly detection, graph/link analysis, unsupervised clustering, and behavioral modeling
- Monitor models for drift, implement automated feedback loops for improvement
- Apply generative and agentic AI to classify unstructured signals, explain outputs, and automate investigations
- Run experiments with defendable conclusions to set operating thresholds and actions
- Collaborate with Policy/Investigations to translate signals into enforcement and use investigator feedback for improved labels
- Represent AI science in cross-functional reviews, linking model decisions to business outcomes
Key requirements- MS/PhD in a quantitative field; 4+ years of production ML experience
- Expert Python and SQL; strong ML/DL framework experience (scikit-learn, XGBoost, PyTorch or TensorFlow, pandas, NumPy)
- Proven track record shipping production ML models and maintaining them in production
- Solid foundation in statistical modeling (classification, regression, clustering, anomaly detection, neural networks, tree ensembles)
- Experience handling class imbalance, noisy labels, and evaluation beyond accuracy (precision-recall, thresholds, cost-sensitive metrics)
- Proficiency with large-scale data ecosystems (Spark/SparkSQL, Databricks, Hive) and Linux
- Ability to explain technical trade-offs to both technical and non-technical audiences and tie performance to business outcomes
- strong communicator
- cross-functional collaboration
- problem framing and critical thinking
- fraud/risk/abuse domain knowledge (preferred)
- graph-based methods and link analysis
- real-time model serving and feature parity
