Lewis Wood
AI Specialist — Production AI Systems
About
I am an AI Specialist at Global Drone Solutions and a recent Computer Science graduate from the University of Western Australia (UWA). My work focuses on production AI systems, particularly AI voice agents, intelligent automation, and real-time communication systems. My projects range from cloud-deployed AI agents, CLI tools and full-stack applications.
I am passionate about building AI systems that solve real-world problems, particularly in automation and monitoring.
Selected Projects
OpenMic — AI meeting notetaker
Feb 2026 · Personal Project
A Python CLI tool that transcribes meetings with speaker diarisation, generates structured notes from customisable templates, and lets you query across your full meeting history in natural language.
Built on ElevenLabs for transcription and LangChain for agent routing across the transcription, note-generation, and query workflows. The terminal UI uses a command-driven interface (/start, /stop, /history, /query, /notes) with templated summary formats for technical, architectural, and decision-focused notes.
Python · ElevenLabs API · LangChain · CLI
Azure MCP Web Scraper
Jun 2025 · Global Drone Solutions
A FastMCP-based web scraping service deployed on Azure Functions, giving a production AI voice agent real-time access to 20+ pages of current website data during live customer calls.
Implements the Model Context Protocol for standardised AI-to-scraper communication, with a content-cleaning and LLM-optimisation pipeline delivering 2–3 second response times. Containerised and deployed on the Azure Functions Consumption Plan with basic authentication — eliminating manual knowledge base updates through on-demand retrieval.
Python · FastMCP · Azure Functions · Docker
Testing AI Models for Road Safety
Aug – Nov 2024 · University Capstone Project
My first real dive into working with AI systems. We built a Python platform for the WA Centre for Road Safety Research that tested how well AI models could interpret road imagery under challenging conditions — fog, graffiti, poor lighting, and explicit content.
Integrated OpenAI's GPT, Anthropic's Claude, and Google's Gemini into a CLI tool that batch processed images and videos with simulated interference, then compared how each model interpreted what it saw.
The client was pleased, but the results were eye-opening. LLM vision capabilities at the time were quite limited — these models, fundamentally built for language, struggled with visual understanding in ways that weren't immediately obvious. This taught me that working with AI isn't just about integration; understanding what these systems fundamentally can and can't do is just as important as knowing how to use them.
Python · OpenAI API · Anthropic API · Google Generative AI · Image processing
Education
University of Western Australia
Bachelor of Science (Computer Science) · Feb 2021 – Dec 2024
Outside of Work
In my spare time, I experiment with local AI model deployment, 3D printing projects, and customise my Linux distribution and my Raspberry Pi.
Get in Touch
I enjoy connecting with likeminded professionals. If you share similar interests or would like to reach out, the quickest ways are below.