Machine Learning Engineer

The French Sourcer; Madrid; On-site; EUR 55,000 - 75,000; Full time

Benefits offered by this job

Health insurance

Job summary

The French Sourcer in Madrid/Barcelona is seeking a Machine Learning Engineer to design and deploy fraud-detection models for real-time decisions. You will build features from transactional data, monitor drifting models and latency, and collaborate with risk, data engineering, platform, and product teams to balance detection rates with customer experience.

Ideal candidates have 3+ years in applied ML, strong Python skills, and experience deploying models in production using CI/CD pipelines.

Qualifications

• 3+ years of experience building applied ML models.
• Strong Python skills and solid software engineering practices.
• Experience deploying and operating models in production.
• Experience with classification, anomaly detection, or risk-scoring methods.
• Experience with imbalanced datasets and evaluation metrics.
• Understanding precision/recall and business trade-offs.

Responsibilities

• Build and maintain fraud detection and risk-scoring models for live transactions.
• Develop features from transactional, behavioural, account, and device data.
• Train and evaluate models against evolving fraud patterns.
• Deploy models into low-latency production systems.
• Reduce false positives while maintaining detection rates.
• Collaborate with risk and engineering to translate fraud scenarios into features.
• Design feedback loops using fraud cases and outcomes.
• Monitor model performance, drift, latency, and distribution.

Skills

Python

Production deployment

CI/CD

Model monitoring

Low-latency inference

Imbalanced data handling

Git and code reviews

Experiment tracking

Tools

Git

Job description

Detect fraud in milliseconds, at the moment a transaction happens, without blocking legitimate customers.

Machine Learning Engineer - Spain

Madrid or Barcelona, Spain · Permanent · Hybrid

What you'd actually work on

• Building and maintaining fraud detection and risk-scoring models used on live transactions
• Developing features from transactional, behavioural, account, and device data
• Training and evaluating models against new and evolving fraud patterns
• Deploying models into low-latency production systems
• Reducing false positives while maintaining effective fraud detection rates
• Working with risk specialists to translate fraud scenarios and business rules into model features
• Designing feedback loops using confirmed fraud cases, manual reviews, and transaction outcomes
• Monitoring model performance, feature quality, drift, latency, and prediction distributions
• Investigating model degradation and changes in customer or fraud behaviour
• Improving model deployment, versioning, retraining, and rollback processes
• Documenting model behaviour and decisions for engineers, risk teams, and auditors
• Contributing to code reviews, automated testing, CI/CD, and ML engineering standards

Where it gets technically interesting

• Real-time inference under strict latency constraints, with decisions required before transactions are completed
• Highly imbalanced datasets where confirmed fraud represents only a small proportion of all transactions
• Fraud patterns that change deliberately in response to existing detection methods
• Managing delayed or incomplete labels when transaction outcomes are not immediately known
• Balancing fraud detection rates against the commercial and customer impact of false positives
• Identifying drift in models and features before it results in significant financial losses
• Combining machine learning outputs with business rules and manual risk controls
• Meeting explainability and traceability requirements for decisions that may need to be reviewed later
• Rolling out new models safely through controlled testing, monitoring, and rollback mechanisms

What we're looking for

• 3+ years of experience developing applied machine learning models
• Strong Python skills and good software engineering practices
• Experience deploying and operating models in production
• Knowledge of classification, anomaly detection, or risk-scoring methods
• Experience working with imbalanced datasets and appropriate evaluation metrics
• Understanding of precision, recall, false-positive rates, and the business trade-offs between them
• Experience with model monitoring, drift detection, versioning, and retraining
• Ability to work with large transactional or behavioural datasets
• Experience with low-latency inference systems
• Confidence working with risk, data engineering, platform, and product teams
• Experience using Git, code reviews, automated testing, and CI/CD

Previous experience in fraud, payments, insurance, credit risk, or anomaly detection would be valuable, but it is not required if you have worked on comparable production machine learning problems.

The company

A fintech or insurtech scale-up operating across Southern Europe, with several hundred employees and high daily transaction volumes.

The machine learning team works closely with risk and engineering to improve fraud detection while limiting unnecessary friction for legitimate customers.

• Health insurance, flexible working, and an equity plan.

Languages: Native or bilingual Spanish and professional English.