Building production AI for digital payments, transaction-graph analytics, temporal machine learning, Generative AI and GPU-accelerated computing.
I am Head of Data Science and an AI leader with 20+ years of experience translating research into scalable, production-ready systems. I specialise in Graph AI, temporal learning, fraud intelligence, money-mule detection and transaction-graph analytics for large digital ecosystems. My work combines AI leadership, applied research and hands-on engineering across production machine learning, scalable analytics and responsible AI.
- 20+ years across AI, data science, machine learning and analytics leadership.
- Lead AI and data science initiatives for one of the world's largest real-time digital payments ecosystems.
- Built production AI systems for fraud intelligence, money-mule detection, anomaly detection, graph analytics, federated AI and synthetic data.
- Published peer-reviewed research in Graph AI, temporal transaction graphs, adaptive fraud detection and high-performance analytics.
- Author of Machine Learning for Finance and educator to 21,000+ learners.
Featured in the Global Fintech Fest 2025 AI Report for national-scale Graph AI applications in money-mule detection, with an attributed expert perspective on model drift, retraining and false-positive reduction. See printed page 11.
- Graph AI and financial crime: graph machine learning, graph neural networks, temporal graphs, transaction-network analysis, fraud detection and money-mule detection
- Scalable production AI: production machine learning, real-time analytics, MLOps, LLMOps, observability, testing and responsible AI governance
- GPU-accelerated analytics: CUDA, NVIDIA RAPIDS, cuGraph and high-performance graph computing
- Generative and agentic AI: LLMs, retrieval-augmented generation, agentic workflows and enterprise GenAI
- Applied machine learning: anomaly detection, time-series forecasting, incremental learning, reinforcement learning and knowledge distillation
| Repository | Focus and Differentiation |
|---|---|
| Topology-Aware Temporal Graph Learning | Reference implementation for temporal node classification using topology-aware features, chronological evaluation and deterministic synthetic transaction graphs. |
| Time-Series Forecasting | Reproducible forecasting benchmark with walk-forward validation, leakage-safe backtesting, classical models, lag-based machine learning, tests and CI. |
| Outlier and Anomaly Detection | Reproducible tutorials and benchmarks covering statistical, distance-based, density-based, isolation, kernel, ensemble and autoencoder methods. |
| Cross-Modal Knowledge Distillation | ANN-to-SNN knowledge distillation for imbalanced tabular classification using spike encoding and hybrid distillation losses. |
- RAPIDS cuGraph PR #5584 — open contribution proposing multi-seed
ego_graphoffset handling in the Python API, with regression tests and backward-compatibility validation.
- Completed three peer reviews for scholarly journals, with the reviewing activity recorded on my Web of Science Researcher Profile.
- Reviewer for the Journal of Advances in Information Technology (JAIT), covering research in multimodal reasoning, retrieval-augmented generation, knowledge graphs and prompt optimisation.
My research focuses on graph machine learning, temporal transaction graphs, fraud intelligence, scalable AI, incremental learning, knowledge distillation, anomaly detection, reinforcement learning and high-performance computing.
Explore the GRADES-NDA 2026 and IEEE ICDE 2026 research highlights →
IEEE International Conference on Big Data (IEEE BigData 2025) · IEEE
DOI: 10.1109/BigData66926.2025.11402449
Proposes an adaptive fraud detection framework combining meta-learning, Kolmogorov–Arnold Networks (KAN) and ensemble learning to improve generalization against evolving fraud patterns in financial transaction systems.
Research Areas: Fraud Detection · Financial AI · Meta-Learning · KAN · Ensemble Learning
IEEE 32nd International Conference on High Performance Computing, Data and Analytics Workshops (HiPCW 2025) · IEEE
DOI: 10.1109/HiPCW66559.2025.00053
Presents a scalable framework for temporal graph motif mining over large-scale financial transaction networks, enabling efficient discovery of transaction patterns for anti-money laundering (AML) investigations and graph intelligence. Evaluated on public benchmark graphs and billion-scale UPI transaction data.
Research Areas: Graph AI · Temporal Graphs · AML · Transaction Intelligence · High-Performance Computing
Selected examples of how my published work has been independently reviewed, cited and extended by international researchers across healthcare simulation and natural-language processing.
View detailed research-impact evidence →
More research: Google Scholar · ORCID · Scopus · Web of Science · ResearchGate · OpenReview · DBLP · Semantic Scholar · ACM Digital Library
- Book: Authored Machine Learning for Finance: Beginner's Guide to Explore Machine Learning in Banking and Finance, published by BPB Publications in 2021. View book overview and author contribution →
- Course: Created Data Analysis for Business and Finance, reaching 21,000+ learners across statistics, probability, regression and time-series analysis. View course and educational impact →
- Technical articles: Explore selected articles published on Towards Data Science, HackerNoon and KDnuggets →
- Quora answers: Explore selected educational answers on AI, probability and machine learning →
I welcome conversations around Graph AI, financial-crime intelligence, scalable machine learning, applied research and responsible production AI. Connect with me on LinkedIn for research collaboration, technical discussions and industry knowledge exchange.