Data Scientist

We are building our Intelligence Core — a new predictive analytics capability at the heart of our sales and client management operations. Instead of relying on gut feel and manual processes, we want every client decision to be informed by data: who is at risk of leaving, who is ready to grow, and where should we focus our attention.

This is a foundational hire. You won't be joining an established data science team and picking up tickets. You will be designing the models, defining the problems, and shaping how the business uses predictive intelligence. You will work closely with the Head of Sales AI Transformation, the Analytics & Sales Controlling team, and front-line sales managers who will act on what you build.

Responsibilities: 

Define and build predictive models from scratch, starting with:

• Churn prediction — identify clients at risk of reducing activity or withdrawing assets, early enough for the sales team to act
• Upsell / cross-sell propensity — score clients by their likelihood to increase assets under custody, and surface the right opportunities
• Client Lifetime Value (CLV) — estimate forward-looking client value to guide resource allocation and relationship management prioritization

Own the full modeling lifecycle:

• Work with raw trading, transactional, and behavioral data from our data warehouse
• Define target variables and operationalize business concepts (e.g., what constitutes "churn" in a brokerage context) into measurable ML targets
• Engineer features from client activity, trading patterns, market conditions, and engagement signals
• Select, train, validate, and iterate on models — starting simple, increasing complexity where it earns its keep
• Design monitoring for model performance, data drift, and degradation over time
• Deliver daily client-level scores that integrate into CRM workflows and sales processes

Bridge data science and business:

• Translate model outputs into actionable insights for non-technical sales managers
• Work with sales leadership to design interventions around model predictions
• Present results, assumptions, limitations, and recommendations to senior stakeholders

Requirments:

Must have:

• 4+ years of hands-on experience building and deploying predictive models on real business problems (classification, regression, scoring)
• Strong proficiency in Python (pandas, scikit-learn, XGBoost/LightGBM/CatBoost) and SQL
• Demonstrated ability to independently frame ambiguous business problems as ML tasks — define the target, engineer the features, choose the approach
• Experience with tabular data at scale: feature engineering, handling class imbalance, temporal validation, avoiding data leakage
• Ability to communicate model results to non-technical stakeholders in plain, actionable language
• Experience working with time-series or event-based behavioral data

Strong advantage:

• Experience with churn prediction, propensity modeling, CLV, or customer scoring in any industry
• Familiarity with survival analysis (Cox proportional hazards, time-to-event modeling)
• Experience with model monitoring in production: data drift detection, retraining pipelines, champion-challenger frameworks
• Background in financial services, brokerage, or fintech
• Experience with probabilistic models for CLV (BG/NBD, Pareto/NBD, Gamma-Gamma)
• Familiarity with SHAP, LIME, or other model interpretability techniques
• Experience with data warehousing tools (BigQuery, Databricks, or similar)

Mindset we value:

• You propose solutions before being asked. When faced with an ambiguous problem, your instinct is to define it and start building, not to wait for a specification.
• You start simple and earn complexity. Logistic regression before gradient boosting. A working baseline before a perfect model.
• You think about what happens after the model is trained. Who uses the output? How do they act on it? Is the score delivered early enough to matter?
• You are comfortable saying "I don't know yet, but here's how I'd find out.