Senior Graph Data Scientist / ML Engineer – Fraud & Forensics

Role Overview

We are seeking an experienced Senior Graph Data Scientist / ML Engineer to support the modernisation and enhancement of a production fraud-detection environment within a large-scale fintech, payments, or mobile-money organisation.

The successful candidate will combine strong expertise in graph analytics, fraud modelling, Python, machine learning, and MLOps. The role will involve enhancing graph-based fraud-detection models while also contributing to the modernisation of the underlying production architecture, including containerisation, orchestration, monitoring, scalability, and performance optimisation.

Key Responsibilities

• Design, develop, and enhance machine-learning fraud-detection models operating in a production environment.
• Develop advanced graph-based fraud analytics to identify relationships and suspicious behavioural patterns across transactions, customers, accounts, and other entities.
• Identify and analyse complex fraud patterns, including:
  • Funnel and aggregator activity
  • Layering
  • Collusion and fraud communities
  • Linked entities
  • Off-ramp tracing
  • Complex transaction networks
• Work hands-on with graph technologies such as NetworkX, Neo4j, or equivalent platforms.
• Develop and maintain production-grade Python-based ML pipelines.
• Consolidate multiple machine-learning models onto shared production infrastructure.
• Containerise ML model pipelines using Docker, while maintaining appropriate isolation between models and environments.
• Implement and enhance workflow orchestration using Apache Airflow or equivalent technologies.
• Design and manage scheduling, sequencing, dependency management, retry mechanisms, and failure isolation.
• Optimise infrastructure and resource allocation to support multiple models running concurrently.
• Conduct load, concurrency, scalability, and performance testing.
• Measure and improve fraud-model performance using metrics such as precision, recall, F1 score, and false-positive rate.
• Develop explainable model outputs suitable for fraud investigators, including risk drivers, linked entities, graph evidence, and alert rationale.
• Implement production monitoring, alerting, pipeline-health visibility, and failure-handling mechanisms.
• Produce technical architecture, deployment, and operational documentation.
• Support testing, deployment, user acceptance, and operational handover.
• Work securely with sensitive financial, customer, and transactional data within the required environment.

Required Experience

Candidates should have strong hands-on experience across Data Science, Machine Learning Engineering, Graph Analytics, and MLOps, including:

• Strong commercial experience as a Senior Data Scientist, Graph Data Scientist, ML Engineer, or similar role.
• Advanced Python development skills.
• Proven experience building, deploying, and supporting production machine-learning models.
• Strong hands-on experience in graph analytics / graph data science.
• Experience with NetworkX, Neo4j, or comparable graph technologies.
• Demonstrable experience developing fraud-detection or financial-crime analytics models.
• Strong understanding of transaction-based and network-based fraud detection.
• Experience with Docker and containerised ML workloads.
• Experience with Apache Airflow or equivalent workflow orchestration technologies.
• Experience operating multiple ML models and pipelines in a production environment.
• Understanding of model monitoring, failure handling, retry mechanisms, alerting, and production support.
• Experience measuring, evaluating, and improving machine-learning model performance.
• Strong understanding of ML explainability and the ability to translate model outputs into actionable information.
• Experience with architecture and technical design documentation.
• Understanding of model versioning, deployment, rollback, and release strategies.

Highly Desirable Experience

• Previous experience within fintech, banking, payments, financial services, or mobile money.
• Experience analysing large-scale financial transaction datasets.
• Experience in Fraud & Forensics, Financial Crime, AML, or transaction monitoring.
• Experience deploying ML solutions within highly regulated environments.
• Cloud migration or cloud-based machine-learning architecture experience.
• Experience with scalable production ML infrastructure.
• Knowledge of data protection, information security, and regulatory requirements applicable to sensitive financial information.

Key Technologies

Essential / Core

Python | Machine Learning | Graph Analytics | NetworkX / Neo4j | Fraud Detection | Docker | Apache Airflow / Orchestration | ML Pipelines | MLOps

Additional / Advantageous

SQL | Git | CI/CD | Kubernetes | Cloud Platforms | Model Monitoring | Graph Databases | Data Engineering

Ideal Candidate Profile

This role is not suited to a pure Data Scientist.

The ideal candidate should have a strong combination of graph analytics, fraud modelling, machine learning engineering, and production MLOps experience.

The candidate should be able to demonstrate that they have personally:

• Built and developed graph-based fraud models using Python or comparable technologies.
• Worked with graph data and transaction networks to identify fraud patterns.
• Built or deployed machine-learning models into production.
• Developed and maintained production ML pipelines.
• Worked with Docker and workflow orchestration tools such as Airflow.
• Supported the infrastructure and operational processes required to run multiple ML models reliably in production.
• Monitored and improved model performance.
• Produced explainable fraud insights that can be used by investigators or operational teams.

A candidate who has excellent Neo4j/NetworkX experience but has never productionised machine-learning models would not be sufficient.

Similarly, a strong MLOps/ML Engineer with extensive Docker, Airflow, and production infrastructure experience but without meaningful hands-on graph analytics and fraud-modelling experience would also not be sufficient.

The ideal candidate is someone who can bridge the gap between advanced graph-based fraud analytics and reliable production ML engineering.