11/12/2026 updated
100 % available
Data Scientist | Machine Learning & Fraud Detection Expert
Ellon, United Kingdom
Worldwide
BSc (Hons) Artificial Intelligence & Computer Science, 2:1 - University of EdinburghAbout me
I build detection systems for things designed to blend in: bot farms posing as listeners, AI music resembling human work, and songs identified from 3 seconds of broadcast audio. I bring fraud investigation, graph analysis and audio AI together, with care for costly false positives.
AI AgentsGCPData AnalysisData IntelligenceMachine LearningTensorFlowData ScienceRandom ForestDeep LearningScikit LearnMachine Learning OperationsArtificial Intelligence Markup Language (AIML)
Machine Learning & Deep Learning:
Expertise in building high-precision machine learning and deep learning models using Python, PyTorch, TensorFlow, scikit-learn, XGBoost, CNNs, and Transformers. Proven track record of developing fraud detection models scoring 500M monthly active users at 99% precision and AI music detection systems with false-positive rates below 0.1%.
MLOps & Cloud Infrastructure:
Strong command of MLOps practices and cloud platforms including GCP (Vertex AI, Cloud Run), AWS, Docker, MLflow, and multi-GPU training. Experience designing end-to-end data-to-model pipelines with dataset versioning, leakage-safe splits, data caching, and batch inference, reducing GPU training costs by 12x.
Graph Analytics & Anomaly Detection:
Applied graph analytics and community detection algorithms (Louvain, Leiden) to large-scale user and device graphs covering over 125M users. Combined with XGBoost, Random Forest, and neural models to uncover bot farms and coordinated fraud at platform scale.
Audio Fingerprinting & Content ID:
Development of an audio fingerprinting system in Python and C++ using ONNX Runtime and FAISS, trained with triplet loss and broadcast-simulation augmentation, achieving up to 98% accuracy on 3-second clips and designed for 150M songs.
SQL & Data Engineering:
Proficiency in SQL, BigQuery, Postgres, and pandas for data engineering and analytics. Ownership of company-wide Looker and SQL dashboards, with experience in ETL pipelines, data caching, and reporting best practices.
A/B Testing & Statistical Experimentation:
Design and execution of live A/B tests for model variants and fraud interventions, applying statistical significance testing to balance fraud detection against genuine user impact before production deployment.
Production Monitoring & Model Drift Detection:
Implementation of real-time monitoring and alerting systems for data ingestion, model outputs, and model drift. Setup and ownership of scheduled model retraining pipelines to maintain production performance.
Temporal Graph Modelling:
Building of temporal graph models using Raphtory, an open-source graph analytics engine, during a software engineering internship at Pometry.
LLM APIs & Model Evaluation:
Familiarity with LLM APIs, model evaluation frameworks, and statistics for assessing and improving machine learning model performance across various domains.
Experience
Professional experience0 - 5 years
Freelancer since2026
Languages
EnglishNative speaker
Project history
Promoted from Data Analyst to Data Scientist I (2024) and II (2025) at a streaming-fraud detection company. Built and led a user-level fraud detection model scoring 500M monthly active users flagging 2.9B suspicious streams at 99% precision, saving clients $1.7M. Designed and built the company's AI music detection capability using CNN and transformer models in PyTorch. Led development of an audio fingerprinting system in Python and C++. Applied graph analytics and community detection to user and device graphs over 125M users. Owned fraud models for major labels and streaming platforms. Designed ML pipelines on GCP and implemented real-time production monitoring. Set technical direction for a small team through mentoring, code reviews, and roadmap planning.
Built temporal graph models with Raphtory, Pometry's open-source graph analytics engine.