Data Scientist (Fraud)

Build predictive defences: Use data and models to support the development of risk mitigation strategies and interventions while preserving and improving the user experience, building, training, and iterating on predictive models that stand up to real adversaries at Kasada scale.

Evaluate rigorously: Pressure test your own work before anyone else has to. Evaluate model performance and trade-offs carefully, so detection decisions stand up to scrutiny from engineering, product, and security operations long after the model goes live.

Partner with engineering, research, and product: Work alongside engineers, researchers, and product managers to take models from notebook to production, integrating them into Kasada's platform and iterating on them as customer needs and attacker behaviour shift.

Make models legible: Make your models and their outputs understandable to both technical and non-technical colleagues. Detection decisions should never be a black box to the teams that rely on them, and that bar sits with you.

Hunt adversarial patterns: Analyse large datasets to surface anomalies, identify adversarial behaviours, and flag emerging attack patterns. Stay current with developments in adversarial ML and cybersecurity, and apply relevant techniques to strengthen our defence capabilities.

You're genuinely curious about fraud and the adversarial landscape, and you ask the right questions about attacker behaviour, false positives, and the real-world impact your models have on legitimate users

You enjoy partnering across engineering, research, and product, and you know how to explain models and their limitations to teammates and stakeholders who don't share your technical background

You have 2+ years of professional experience in data science or applied ML, with a solid foundation in statistical concepts, sampling, time-series data, and hands-on predictive modelling (e.g., gradient boosted trees, random forests, deep learning)

Proficiency in Python, SQL, and standard ML libraries (e.g., scikit-learn, PyTorch, TensorFlow)

Experience evaluating predictive models in production or production-like settings, including thinking through precision, recall, calibration, and how models behave as attackers adapt

Strong problem-solving and analytical skills, with a keen attention to detail and a bias toward pressure testing your own work before shipping

Experience working within cloud environments (like AWS)

AWS, Clickhouse

Python, scikit-learn, PyTorch, TensorFlow