Senior Data Scientist, ML— Fraud Detection & Effectiveness

Adobe•Washington, DC

2d•$133,100 - $236,400•Hybrid

About The Position

We are seeking an experienced Senior Machine Learning Data Scientist to build fraud and abuse detection models and measure how effectively they work. This role combines hands-on modeling with deep experimentation, evaluation, and analytics to improve detection and quantify business impact. You will work across the fraud lifecycle — from modeling and ground-truth definition to performance measurement, monitoring, and executive-ready insights!

Requirements

8+ years in applied Data Science / ML, with experience building and evaluating production ML models.
Strong foundation in statistical and classical ML, experimentation, model evaluation, and performance measurement.
Strong hands-on Python and SQL skills working with large, complex datasets.
Experience with model monitoring, drift, false-positive/false-negative analysis, and imperfect or delayed labels.
Strong data visualization and storytelling skills — able to translate complex analysis into clear insights and recommendations.
Strong analytical judgment, ownership, and ability to operate independently through ambiguity.
Bachelor's or equivalent experience in Statistics, Mathematics, Computer Science, or related field; advanced degree a plus.

Nice To Haves

Experience in fraud, abuse, risk, identity, trust & safety, or other adversarial domains.
Experience with anomaly detection, clustering, behavioral modeling, or prevalence estimation.
Experience with labeling frameworks, weak supervision, active learning, or human-review systems.
Familiarity with LLMs and AI-assisted evaluation/analysis.
Experience evaluating multi-layered risk controls and automated decisioning systems.

Responsibilities

Build and tune ML models for fraud and abuse detection using statistical and classical ML techniques.
Develop robust evaluation frameworks, datasets, and metrics to measure model and mitigation effectiveness.
Analyze false positives/negatives, model drift, and emerging fraud patterns to continuously improve detection.
Define ground truth, labeling approaches, and fraud taxonomies that support reliable model development and evaluation.
Design experiments and evaluate tradeoffs across precision, recall, customer impact, and fraud loss.
Build dashboards and metrics that translate detection performance into measurable business impact.
Pressure-test models and data for leakage, bias, data-quality issues, and other sources of misleading results.
Partner across engineering, product, policy, and risk teams to turn insights into detection improvements and business decisions.