Senior AI Scientist

Intuit's Trust & Safety organization protects millions of customers and the money, identities, and data they trust us with. Fraud is an adversarial problem: the actors behind malicious activities such as account takeover and product abuse change tactics swiftly and unpredictably. We are looking for a Senior AI Scientist to build the detection models that stay one step ahead of them.

You will own fraud detection models end to end, from problem framing and data discovery through feature engineering, production deployment, and monitoring in a live environment. The work spans real-time risk scoring, entity-level and graph-based anomaly detection, and generative and agentic AI applied both to catching bad actors and accelerating the teams who act on our signals. You will join a distributed team of AI scientists and partner closely with other teams in Trust & Safety (Policy, Investigations, ML Engineering, and Analytics) to design and release models that excel at catching fraud while minimizing customer friction and operational cost.

Responsibilities

• Design, build, and deploy machine learning models that detect fraud and abuse across the customer lifecycle (account creation, authentication, and in-product activity) in both real-time and batch settings.
• Own models end to end: discover data sources, build ETL, engineer features, train and validate, partner with AI Engineering to productionize, then monitor, retrain, and improve in production.
• Develop detection approaches beyond standard supervised classification, including entity-level anomaly detection, graph and link analysis, unsupervised clustering for novel attack patterns, and behavioral modeling.
• Monitor deployed models for drift and emerging attack patterns, and build the feedback loops that allow for quick and automated model retraining and improvement.
• Apply generative and agentic AI to complement classical ML tools: classifying unstructured signals, explaining model outputs and decisions, and automating manual investigation workflows.
• Design and run experiments and champion/challenger tests, drawing defensible conclusions to determine operating thresholds and downstream actions.
• Partner with Policy and Investigations to turn detection signal into enforcement action, and investigator feedback into better features and labels.
• Represent AI Science in cross-functional reviews, translating model design decisions and performance into tangible business implications for policy, product, engineering, and compliance stakeholders.

Qualifications

Required

• MS or PhD in Computer Science, Statistics, Applied Mathematics, Operations Research, Physics, or a related quantitative discipline.
• 4+ years of industry experience building and deploying machine learning models in production (fintech, consumer tech, security, risk, or e-commerce preferred).
• Expert proficiency in Python and SQL, with deep experience in modern ML/DL frameworks (scikit-learn, XGBoost or equivalent gradient boosting, PyTorch or TensorFlow, pandas, NumPy).
• Demonstrated experience with models that made real decisions in production: you have owned something that shipped, watched it degrade, and fixed it.
• Solid foundation in statistical modeling and ML: classification, regression, clustering, anomaly detection, neural networks, and tree ensembles.
• Experience handling severe class imbalance, delayed or noisy labels, and evaluation beyond accuracy: precision-recall trade-offs, threshold selection, cost-sensitive metrics.
• Proficiency with large-scale data ecosystems (Spark/SparkSQL, Databricks, Hive, or equivalent) and comfort in a Linux environment.
• Demonstrated ability to explain complex technical concepts and trade-offs to both technical and non-technical audiences, and to link model performance to customer and business outcomes.

Preferred

• Direct experience in fraud, risk, abuse, security, anti-money-laundering, or another adversarial modeling domain.
• Graph-based methods for entity resolution or ring detection: link analysis, belief propagation, graph neural networks, or community detection.
• Real-time model serving, including latency-constrained feature computation and online/offline feature parity.
• Applied LLM experience in production: classification over unstructured text, evaluation methodology, or agentic patterns such as tool calling and multi-step reasoning.
• Familiarity with feature stores, model monitoring, and MLOps practice for models requiring frequent retraining.
• Experience working with investigations, operations, or policy teams where model output drives human review and enforcement.

Intuit provides a competitive compensation package with a strong pay for performance rewards approach. This position may be eligible for a cash bonus, equity rewards and benefits, in accordance with our applicable plans and programs (see more about our compensation and benefits at Intuit®: Careers | Benefits). Pay offered is based on factors such as job-related knowledge, skills, experience, and work location. To drive ongoing fair pay for employees, Intuit conducts regular comparisons across categories of ethnicity and gender.

San Diego $165,500 - $223,500