London, England, United Kingdom
- British citizen with more than 2 decades experience and expertise in quantitative analytics, machine learning, data science and data engineering - applicable to the banking, insurance and consulting domains. - Extensive expertise in deep learning, Bayesian inference, Natural Language Processing (NLP), Computer Vision (CV), etc., applied to various use cases such as Asset Liability Management (ALM), actuarial pricing, trader surveillance, anti-financial crime, etc.. - MSc in Applicable Mathematics from the London School of Economics and Political Science (LSE), UK. - BSc in Economics (majored in Econometrics and Mathematical Economics) from the University of Manchester, UK. - Invited guest lecturer at the University of Oxford Department of Continuing Education (DCE) course "Artificial Intelligence - Cloud and Edge Implementations". - Personal GitHub profile: https://github.com/hamaadshah - Personal GitHub repositories on data science have 350+ stars & forks. - CFA levels 1 and 2 exams passed on first attempts with CFA level 3 exam to be completed in due course.
- Data engineering and data science contractor in Global Financial Crimes Division (GFCD): Effectively working at the principal / Vice President (VP) level of the firm. - Leading the design, tuning, and optimisation of global transactions monitoring and sanctions screening models, enhancing detection effectiveness, segmentation logic, and alert quality across MUFG’s GFCD. - Oversee model validation, scenario calibration, and analytical governance, ensuring regulatory alignment with FCA, OFSI, FATF, and EU expectations. - Manage and mentor a specialist analytics team, driving cross‑regional collaboration to deliver consistent, data‑driven improvements in AML and sanctions risk coverage. - Developed data science models, especially unsupervised, self‑supervised and semi-supervised learning, given we have a lot of unlabelled data however only a few genuine true alerts, to optimise detection performance and reduce false positives and false negatives. - Leveraging Databricks, Snowflake, Python, Spark, Snowpark, Pandas, Polars and PyTorch to build scalable analytics pipelines and model‑driven tuning frameworks. - Partnering with compliance, technology, and operations teams, translating complex analytical insights into clear recommendations for senior management and regulators.
- Contractor in the trade compliance team: Effectively working at the Senior Principal level of the firm (the highest level for an IC – parallel track to management level however seniority comparable to VP / SVP, etc). - Designed and deployed a time series forecasting pipeline in Azure Databricks (Spark) using TabNet and LightGBM to predict commodity unit values for trade compliance. - Engineered fact and SCD2 tables from raw bronze layer transactions and trades data, writing to a Delta Lakehouse (Azure Blob external storage) to support multiple downstream projects. - Built 1,000+ engineered features at terabyte scale using Spark and Polars within RDD partitions, capturing nuanced historical value per unit patterns for predictive modelling. - Developed a hierarchical HS code classification system by fine tuning context-aware BERT embeddings with a LightGBM output layer, achieving >90% precision and recall across commodity categories. - Applied beam search and dynamic programming to collate top K HS codes per shipment, improving classification robustness. - Published a real time feature store to Azure Cosmos DB, enabling sub millisecond feature retrieval for live scoring in production. - Served models via FastAPI, ensuring scalable and low latency deployment. - Prototyped a Dual Use Goods (DUG) engine using LangChain / AdalFlow frameworks for prompt engineering and Retrieval Augmented Generation (RAG) to identify commodities with potential dual use applications (e.g., electronics usable in both consumer and defence contexts). - Partnered with DevOps and software engineering teams across the full SDLC (dev; UAT; prod), establishing CI/CD pipelines to accelerate deployment cycles. - Collaborated with senior product leaders to translate customer needs into data driven trade compliance solutions, ensuring measurable business value for external stakeholders.
- Contractor within the Financial Crime & Transaction Monitoring team, delivering advanced analytics solutions for anti money laundering and fraud detection. - Designed and implemented an unsupervised machine learning framework with hand crafted feature engineering to surface anomalous transactions and uncover hidden, complex typologies beyond rule based detection. - Built scalable pipelines leveraging Python, PySpark on Kubernetes, ThetaRay, Scikit Learn, PyTorch, TensorFlow, with orchestration via Airflow and experiment tracking in MLflow. - Engineered data infrastructure with MinIO for distributed object storage, ensuring high performance access to large scale transaction datasets. - Partnered with compliance and risk stakeholders to translate technical models into operational surveillance capabilities, strengthening the bank’s financial crime controls.