ML engineer · AI security · Birmingham, UK
I build machine learning that holds up against attackers.
I have a bachelor's in cybersecurity and I'm doing a master's in AI. I build machine learning systems end to end, from the data to the deployed model, and I evaluate them the way a security engineer would: fair baselines, held-out tests, and a clear record of where they still fail.
Would this message hijack an AI assistant?
Pick a message. The highlighted part is the attack.
Real decisions recorded from the gateway on 28 Sep 2026. v0.1 = rules only · v0.3 = rules + my fine-tuned classifier.
Work
Each project is a case study: problem, research, decisions, results, and what I'd change.
LLM Security Gateway
Stops prompt injection in English and Spanish before it reaches an AI model, including attacks hidden inside emails and documents.
56% → 97%attacks caught after adding my fine-tuned classifier to the rules
● Completed · BSc final year projectBlockchain-Enhanced Firewall
A Python firewall that blocks floods, worms and known bad addresses in real time, and records every decision in a tamper-evident hash chain.
6 testscontrolled attack simulations: every attack blocked
Intrusion detection, end to end
A network intrusion detector with the full MLOps loop: data versioning, training pipeline, drift monitoring and retraining.
About
I came to machine learning from security, so I test my own work the way an attacker would. When a number looks too good, I go looking for the shortcut behind it.
In my LLM Security Gateway project, that habit found two bugs in my own dataset: the validation split had no emails, and a model had memorised 25 harmless sentences. Fixing them made the scores lower and the results true.
I publish the datasets, models and notebooks behind every result, so anyone can check them.
Machine learning
Fine-tuning transformers, evaluation design, dataset construction, baselines, error analysis, model export (ONNX)
Security
Threat modelling (OWASP Top 10 for LLMs, MITRE ATLAS), prompt injection, red-team test sets, detection engineering, network security, firewalls (iptables)
Engineering
Python, PyTorch, FastAPI, ONNX Runtime, pytest, GitHub Actions, Hugging Face, Google Colab
Languages
English, Spanish