ML engineer

Purpose of the role Build, train, deploy, and maintain all analytical and machine learning models serving the Investment Division and, in later phases, the Group’s operating divisions. This includes the five core investment model families (factor decomposition, anomaly detection, NLP/document intelligence, portfolio optimisation, manager screening) and operating division models (DuPont automation, credit scoring, demand forecasting). Work directly with portfolio analysts to validate model outputs and iterate based on domain feedback.

Key responsibilities

• Build and deploy factor exposure decomposition models: rolling Fama-French, Barra, and custom factor regressions for equity and fixed income portfolios. Daily batch processing on gold layer return series.
• Build and deploy anomaly detection for manager monitoring: style drift detection (rolling z-scores on factor loadings), return pattern breaks (CUSUM, Bayesian changepoint detection), correlation regime shifts (KS tests).
• Build and deploy the NLP/Document Intelligence Engine models: document classification (LLM fine-tuning on Group document corpus), financial data extraction from PDF/Excel, summarisation and sentiment tracking for fund manager letters, semantic search via RAG on the vector store.
• Build and deploy portfolio optimisation engine: Black-Litterman, risk parity, mean-variance with constraints (Shariah compliance, concentration limits, liquidity requirements). IC-facing rebalancing recommendations.
• Build and deploy manager screening algorithm: composite scoring from eVestment/Mercer data (alpha, factor-adjusted returns, drawdown, correlation with existing managers, fee-adjusted performance).
• Build DuPont automation for KSA equities (40+ stocks) and OMACO (14+ entities): automated five-stage decomposition updated quarterly from financial statement extraction.
• In later phases, build operating division models: credit scoring for Nama Financing, demand forecasting for automotive and electronics, RevPAR analytics for real estate.
• Track all experiments, model versions, and parameters using MLflow. Maintain model documentation for IC audit trail.
• Work directly with portfolio analysts: present model outputs, validate against domain knowledge, iterate based on feedback.
• Monitor deployed models for drift and degradation; retrain and update on a scheduled basis.

Qualifications and experience

• 5–8 years of experience in machine learning engineering, with at least 2 years deploying production ML models.
• Strong proficiency in Python, scikit-learn, statsmodels, PyTorch or TensorFlow, and MLflow.
• Experience with NLP and LLM deployment: fine-tuning open-source models (Llama, Mistral), RAG architectures, vector databases, document processing pipelines.
• Understanding of quantitative finance: factor models, performance attribution, risk analytics, portfolio optimisation. CFA or FRM a plus.
• Experience deploying models on Databricks or equivalent Spark-based platforms.
• Familiarity with financial data (return series, holdings data, fundamental data, fund manager reports).
• Master’s degree in machine learning, statistics, quantitative finance, or equivalent.
• Strong communication skills: ability to explain model outputs and limitations to non-technical investment professionals.