Data Scientist Outlier Detection (40%, 3 months)

Responsibilities

• Consult, support and provide knowledge for the development and roll-out of a global model on market abuse (time series outlier detection system)
• Analyze and justify methodological choices based on data properties, business objective, and operational constraints
• Challenge flawed assumptions, improve model robustness, or prevent false conclusions due to spurious correlations, lookahead bias, or poor experimental design
• Analyze, design and develop in Python using Pandas, NumPy, Scipy, Scikit-learn etc.

Requirements

• Several years of professional experience in quantitative roles applying advanced data science and statistical modelling within the financial industry
• Very good command Python for data science
• Deep hands-on experience with Pandas, NumPy, Scipy, Scikit-learn, statsmodels, and either PyTorch or Tensorflow
• Extensive experience with financial time series data including pre-processing irregular, asynchronous, and microstructurally rich data
• Experience in developing and operationalising time series outlier detection systems designed for monitoring abnormal market behaviour (linking detected anomalies to external market events and and integrating NLP outputs like news sentiment into event-driven detection logic)
• Advanced theoretical and applied knowledge of time series econometrics and anomaly detection methodologies like ARIMA, SARIMA, VAR models, GARCH-family volatility models (incl. EGARCH, GJR-GARCH), change point detection algorithms (PELT, Binary Segmentation), forecast residual analysis with volatility adjustment and unsupervised learning techniques: Isolation Forests, One-Class SVM, DBSCAN, LSTM-based autoencoders
• Strong proficiency in SQL, particularly in Oracle
• Fluent in English

Nice to have

• Experience in regulatorily sensitive banking domains such as market or trade surveillance

Personality

• Strong communication skills
• Able to translate complex technical findings into actionable insights