// GRADUATE AI ENGINEER & ROBOTICS SPECIALIST
MSc Robotics & AI graduate building end-to-end intelligent systems — from a 94.17%-accurate CNN human-detection model to a fully autonomous SLAM navigation stack. Currently a Research Intern at the Eureka Robotics Centre, developing safety-critical NLP for human–robot interaction.
Complete ROS 2 autonomy pipeline — SLAM mapping, A* global and DWA local planning — with a real-time human-detection module for context-aware behaviour, validated across multiple Gazebo scenarios.
ResNet-18 fine-tuned for human/non-human classification on a custom dataset expanded with low-light augmentation; evaluated with precision, recall and F1, and designed for lightweight deployment on mobile robots.
Safety middleware that confidence-scores NLP outputs and applies threshold + fallback logic, so robots default to a safe state instead of executing misrecognised or hallucinated commands. Publication potential.
An open-source MCP routing proxy that lets Claude Desktop call locally-hosted Ollama models and translates Anthropic-style payloads through LiteLLM, keeping model-routing state ready for a future desktop companion UI — with tools to discover local models, select a route, forward prompts to Ollama, and delegate back via sampling.
I came to AI from the systems side. A Computer Science Engineering degree — algorithms, operating systems, embedded C — and a full-stack Python internship taught me to ship software that survives contact with real users and real hardware.
My MSc in Robotics & AI at Cardiff Met put that foundation to work: training and evaluating deep-learning models, building ROS 2 autonomy stacks, and programming NAO humanoids for assistive physiotherapy. Now, at the Eureka Robotics Centre, I research how robots can act safely when their language models are uncertain — the messy, critical space where models meet motors.
Freelance builds using the same engineering rigour I bring to research — semantic markup, measurable performance, and a clean handoff to a live URL.