Financial Crime AI
Transaction monitoring, fraud detection, anomaly detection, behavioural profiling and explainable model design.
I work at the intersection of artificial intelligence, financial crime and quantitative risk — turning complex data into robust, explainable models and better decisions.
Combining statistical rigour with production-minded AI engineering — across finance, regulated modelling and real-world systems.
Transaction monitoring, fraud detection, anomaly detection, behavioural profiling and explainable model design.
PD / LGD / EAD, IFRS 9, Basel frameworks, challenger models, validation and stability monitoring.
Risk factor analysis, Monte Carlo simulation, VaR, scenario design and interest-rate / option modelling.
Scalable data pipelines, experiment tracking, model delivery, computer vision and natural language processing.
A selection of professional domains and published research. Employer work is presented at a high level, without confidential implementations or data.
Anomaly detection, customer-behaviour modelling and transaction-monitoring analytics across high-volume retail financial data.
Risk decisioning, scorecard evaluation, gradient-boosting challengers, portfolio drift and regulatory model governance.
Transformer-based survival analysis for patients with multiple visits, combining time-varying covariates and synthetic data.
Similarity learning through knowledge distillation and quadruplet-loss techniques, with retrieval performance evaluation.
Time-series models using LSTM, Transformers and ensemble methods to predict short- and medium-term city demand.
Diabetic-retinopathy grading using ensembles, knowledge distillation and Grad-CAM for model interpretability.
Professional case studies describe areas of work only; no confidential data, models, code or employer systems are included.
From production AI and backend engineering to credit risk, financial crime and advanced modelling in banking. Detailed responsibilities reflect my CV.
Leading applied AI and statistical detection work in a high-volume financial-crime environment.
Retail risk decisioning and model governance across Santander credit products.
Payment-fraud analytics and applied ML for risk controls.
Delivered client-facing AI products and production data services.
Designed ML-assisted analytics and automation for the electricity sector; recognised as Employee of the Year 2022.

Research and applied AI engineering across payment risk, chatbots and enterprise analytics.
Built industrial AI and backend systems for production and internal logistics.
My academic foundation spans mathematics, statistics, computer science and advanced machine learning. I translate these methods into interpretable, production-oriented systems across financial services and beyond.
Applied research exploring representation learning, explainable computer vision and forecasting. The projects below are described in my CV; linked demos are public research artefacts.
Developed an image-similarity retrieval approach with supervised and unsupervised learning, transfer learning and metric-learning losses, including the proposed quadruplet-loss method. Evaluated retrieval using mean average precision and precision–recall curves.
Investigated ensemble learning, knowledge distillation and pre-trained vision models for retinal-image classification. Applied Grad-CAM to visualise areas contributing to model predictions, alongside data-quality analysis before training.
Built a forecasting approach for electricity demand in Da Nang, Vietnam, covering three-day, one-week and one-month horizons. Used LSTM and Transformer-based time-series models and synthetic data to support model training.
Formal study, funded scholarships, research and technical competitions, from core computer science through advanced AI and statistics.
Academic degrees, achievements and competition results are presented as supplied in the 2026 CV. Publication and demo links point to external research resources.
From modelling research to production engineering and risk governance.
Interested in financial AI, quantitative modelling, responsible machine learning or research collaboration? I'd be glad to connect.
If you choose Allow, the portfolio owner receives one Telegram message when you visit, containing the approximate city, region and country inferred from your network IP address, the date/time, page path, referring website domain (if available) and device category. This does not identify you personally and can be inaccurate, particularly when using VPNs or mobile networks. The site does not store visitor records or send raw IP addresses to Telegram. Vercel handles the connection and derives approximate geography; Telegram processes the notification and retains messages under its own policies. You can decline without affecting the site. Your choice is remembered only in this browser tab. Use the Privacy choices control below to change it for future visits; an alert already sent cannot be recalled. For questions, contact the portfolio owner using the email link above.