$ ls ~/projects
Security tooling built end to end: threat intelligence, phishing defence, offensive automation and AI system security.
Threat intelligence dashboard that ingests worldwide IP abuse reports and surfaces high-risk regions and attack trends. Live API ingestion with a simulation mode for offline analysis.
Security awareness trainer that drills users on real phishing indicators: spoofed sender domains, deceptive display names and urgency cues. Scored rounds with immediate feedback.
Autonomous agent that solves web and reconnaissance CTF challenges inside a sandboxed Docker container, then publishes its own write-up of the exploitation path.
Evaluation harness for retrieval-augmented LLM systems. Injects counterfactual facts into the corpus so a correct answer can only come from retrieval, never pretraining, then scores retrieval rank, answer correctness and groundedness. A separate judge model rules on entailment to avoid self-grading bias.
Data pipeline and dashboard tracking UK fuel prices from public retailer feeds, with cleaning, aggregation and regional comparison.
Monte Carlo simulation of blackjack strategies under weighted deck conditions, used to model expected value across thousands of hands.