Senior Machine Learning Engineer, Ads

Unity Technologies•Mountain View, CA

2d•$140,700 - $232,300

About The Position

Offerwall by Unity is seeking a Senior Machine Learning Engineer to own the machine learning work for Offerwall fraud prevention. The role focuses on taking real-time anomaly detection and device risk models from development into production, building the labeled fraud dataset and evaluation methodology, and modernizing model training, serving, and monitoring. The engineer will work with data engineering, backend, and product partners whose systems the models use, and will have responsibility for fraud machine learning from end to end. The position covers production fraud models, the streaming and batch features that feed a Redis online feature store, the training and serving stack, and investigation of emerging fraud patterns. It also includes monitoring model and feature health, documenting runbooks and data lineage, and participating in on-call for the fraud detection stack.

Requirements

Several years of experience building machine learning models and taking them through production, including deployment, monitoring, and post-launch iteration.
Hands-on experience with fraud detection, abuse prevention, risk scoring, or anomaly detection, including work with sparse, noisy, or adversarial labels.
Experience with low-latency model serving and online/offline feature pipelines, such as Redis or another online feature store, Kafka or similar streaming, and Airflow.
Ability to build labeled datasets from imperfect signals and design evaluation methods that remain useful as fraud patterns shift.
Strong Python and SQL skills.
Working knowledge of a cloud ML platform such as AWS SageMaker, Kubeflow, or KServe.
Ability to take ambiguous problems through to shipped outcomes with limited oversight, document work, collaborate with backend, data, and product partners, and use AI coding tools to work faster.
Sufficient knowledge of English for professional verbal and written communication with colleagues and partners worldwide.

Nice To Haves

Experience with ad fraud, install or attribution fraud, or fraud in rewarded or incentivized advertising.
Experience with unsupervised or semi-supervised detection methods such as autoencoders, isolation forests, or graph-based approaches to device and user linking.
Experience migrating ML serving between platforms or building MLOps tooling on Kubernetes.
Background in mobile ad tech, mobile games, or app monetization.

Responsibilities

Take real-time anomaly detection and device risk models from development to production so model outputs inform payout and blocking decisions.
Create and maintain a golden labeled fraud dataset, and define evaluation methods and baseline fraud-efficacy metrics for assessing model and rule changes.
Design and evolve streaming and batch features that feed the Redis online feature store, partnering with data engineering on freshness, correctness, and lineage.
Plan and drive evolution of the training and serving stack, including evaluating a potential move from SageMaker to Kubeflow and KServe with attention to real-time scoring latency.
Investigate fraud spikes and analyze user and device behavior to develop new detection signals, features, and models.
Monitor production model and feature health, write runbooks and data lineage documentation, and participate in on-call for the fraud detection stack.

Benefits

Comprehensive health, life, and disability insurance
Commute subsidy
Employee stock ownership
Competitive retirement or pension plans
Generous vacation and personal days
Leave and family-care programs supporting new parents
Office food and snacks
Mental health and wellbeing programs and support
Employee Resource Groups
Global Employee Assistance Program
Training and development programs
Volunteering and donation matching program
Potential eligibility for equity awards and company incentive plans, such as annual discretionary bonuses or sales commissions