ML Scientist

We are looking for the first ML research hire on the team. Your job will be to turn our telemetry into new detection and prevention capabilities: from understanding the problem alongside the Security team, to ideating, prototyping, and validating models and algorithms on real data.

We are not looking for someone who applies recipes. We are looking for someone who enjoys tackling an open problem with an adversary that adapts, and who can move comfortably between the paper and the working prototype.

Here is a little taste of your challenge:

•  Work side by side with the Security team to understand current and emerging attack techniques against mobile apps, and translate them into well-framed data problems.
• Explore and analyze our telemetry (datalake on ClickHouse and S3) to uncover signals, patterns, and anomalies with detection value.
• Design, prototype, and evaluate attack detection and prevention models: anomaly detection, classification with scarce labels, behavioral models, event time-series analysis.
• Define how success is measured: detection metrics, acceptable false-positive rates, and evaluation methodology when complete ground truth doesn't exist.
• Take validated prototypes toward production in collaboration with the engineering team, and lay the foundations (features, datasets, labeling processes) for the team's ML capability to grow.
• Stay current with research in ML applied to security (anomaly detection, adversarial ML, fraud) and bring actionable ideas.

Who are you?

• Proven experience (5+ years, or PhD plus applied experience) researching and developing ML models on real-world problems, ideally in security, fraud, anti-bot, anti-malware, or similar adversarial domains.
• Strong command of anomaly detection and unsupervised / semi-supervised learning: labeled data is scarce in this domain, and that shapes the entire approach.
• Experience with large-scale telemetry or event data: logs, time series, user or device behavior data.
• Full autonomy with raw data: solid Python (pandas/polars, scikit-learn, PyTorch or similar), advanced SQL, and comfort working directly against a datalake without prepared datasets.
• Product mindset: the ability to go from a research idea to an evaluable prototype in weeks, not months, and to communicate results to non-technical stakeholders.
• Genuine curiosity about cybersecurity and the drive to learn the domain in depth alongside the team's experts.

Nice to have

• Prior experience in mobile security, EDR, banking anti-fraud, or intrusion detection systems.
• Knowledge of adversarial machine learning (model evasion, robustness).
• Experience with AWS S3, ClickHouse or other high-volume analytical databases.
• Publications, open-source contributions, or talks in applied ML or security.
• Experience deploying models to production (basic MLOps).

Practically Speaking, what's in it for you:

• The chance to define the company's ML strategy from scratch, with direct product impact and full visibility.
• A technically fascinating problem: a real adversary that evolves, massive untapped data, and freedom to research.
• Daily collaboration with a Security team with deep domain expertise.
• Optimized onboarding – for your optimal start and customized induction.
• Technical ownership – take responsibility for key components and shape how we build our platform.
• Enthusiastic team – waiting to help you succeed and collaborate with experienced engineers.
• Beautiful office spaces – in the heart of Barcelona, 200m from Sants Estació with regular team/community events (breakfast, lunch…).
• Good transport connections – easy to reach by public transport.
• Flexible working hours & Hybrid working!

#LI-Hybrid
#LI-LS1