My Projects

Selected work in large-scale ML, generative AI, model optimisation and production systems.

AI Search Visibility platform
LLM Evaluation · RAG

AI Search Visibility

Queries GPT-4o and Gemini through APIs and measures how brands appear in AI-generated answers.

Real-time anomaly detection system
Anomaly Detection · Distillation

Real-Time Anomaly Detection at CERN

Unsupervised anomaly detection on more than one billion events, deployed for continuous use under strict latency limits.

BitNet transformer optimisation for FPGA
LLM · FPGA

BitNet Transformer Optimisation

Converted a 2.08-billion-parameter ternary transformer into lookup-table operations for efficient FPGA inference.

Hardware-aware ML compression project
Model Compression · Hardware

ML Compression with HPE Labs

Combined model distillation with specialised hardware to achieve 12× compression and 24 ns inference latency.

SDXL inference optimisation study
Generative AI · Optimisation

SDXL Inference Optimisation

Benchmarked 17 optimisation configurations and reduced image-generation latency from 5.72 to 0.876 seconds.

OverPrompt data analytics agent
LLM · Data Analytics

OverPrompt Analytics Agent

An LLM assistant that answers questions using deterministic analysis of more than one million structured records.

YOLO distillation for robotics
Computer Vision · Distillation

Robotics Object Detection

Distilled YOLOv8s into a 6.2 MB model that improved mAP50 by 46% over the baseline student and ran 1.8× faster than the teacher.

GitHub anomaly detection pipeline
Anomaly Detection · MLOps

GitHub Anomaly Detection

An end-to-end Isolation Forest pipeline with experiment tracking, scheduled workflows, API inference and drift monitoring.

Credit risk scoring project
Machine Learning

Credit Risk & Behaviour Scoring

End-to-end PD/LGD modelling, behavioural segmentation, anomaly detection and FastAPI deployment.