Senior Data Science Engineer

We are looking for a Senior Data Science Engineer to own an end-to-end fraud detection system, from data profiling and feature engineering to deployment, monitoring, and handover. You will build and maintain production models on high-volume event data, working directly with fraud analysts in a restricted client environment where data remains inside the network.

Required for this role
5+ years of experience in data science or ML engineering, including at least two production models that you deployed and maintained.
Hands-on experience with fraud, anomaly, or abuse detection using high-volume event data, such as payments, telecom, ad fraud, or security logs.
In-depth knowledge of gradient boosting and Isolation Forest, including calibration and threshold tuning for heavily imbalanced datasets.
Experience with weak supervision, learning from noisy labels, and time-windowed feature engineering on streaming or micro-batch data.
Practical knowledge of SHAP explainability and the ability to translate model outputs into clear rules for non-ML analysts.
Strong Python and SQL skills, with end-to-end ownership of data profiling, ingestion, feature pipelines, model development, and production support.
Experience with MLOps on virtual machines, including Docker, MLflow, scheduled pipelines, retraining workflows, and drift monitoring.
Experience working through VPN in restricted client environments where data cannot leave the network, together with clear written English for presenting results directly to fraud analysts.
Knowledge of telecom fraud, SIP signalling, CDRs, E.164 numbering, Wangiri, IRSF, or CLI spoofing, as well as graph features or community detection in caller-callee networks.
Understanding of privacy and regulated model practices, including keyed hashing, pseudonymization, federated learning, GDPR, DPIA work, or regulated model documentation.

Your responsibilities
Lead discovery with the pilot operator, profile the available data, and prepare a clear go/no-go report.
Build the ingestion, tokenization, and time-windowed feature engineering pipeline.
Train, evaluate, calibrate, and validate three fraud detection models on heavily imbalanced event data.
Develop explainability outputs and convert model signals into readable rule suggestions for fraud analysts.
Build the rule suggester, review API, and audit log required for analyst review and traceability.
Package the production deployment, scheduled workflows, drift monitoring, and model retraining pipeline.
Run the shadow pilot together with the client's fraud analysts, review model behavior, and refine thresholds and rules.
Maintain the deployed solution throughout the engagement and complete the technical handover.

What you get
Your time off
18 paid vacation days and 10 paid sick days annually
10 Ukrainian public holidays
Maternity and paternity leaves
Marriage and Parenthood Package
Additional leave for major life events
Sabbatical leave opportunities

Learning & growth
Sombra University workshops and internal learning programs
Tech Communities and knowledge sharing sessions
Language courses and workshops
Mentorship opportunities

Health & well-being
Sports compensation or health insurance coverage
Participation in races and marathons
Corporate doctor (telemedicine)
Well-being initiatives and workshops

And even more
Company-provided technical equipment
PE administration and tax support
Internal referral program
IT Club loyalty program
Company events and volunteering initiatives