Martin Michaux

Building new AI-driven products and working independently on applied data, AI, and energy-related systems.

Worked on applied research around knowledge-graph-based optimisation for RAG systems, while supporting the Data & AI team. • Developed KarGus, a Graph Neural Network trained on a dynamic knowledge graph built using NLP techniques • Explored and prototyped approaches using RAG, prompt engineering, NER fine-tuning, sentiment analysis, and GANs • Participated in hackathons with Google (Princess Máxima Center) and Accenture (Plan International) • Built FurbAI, a Gen-AI companion for children with cancer using GCP tools (text-to-speech, RAG, sentiment analysis)

Worked on AI chatbot development and data infrastructure within the EP&A team. • Built and deployed a RAG chatbot for the Design & Construction team • Designed and maintained BI tools and ETL pipelines used in day-to-day operations • Worked with AWS services including SageMaker, OpenSearch, Bedrock, Airflow, and Glue • Project showcased at AWS re:Invent (Las Vegas)

Actuarial consulting project for a water distribution company. • Built risk and economic valuation models to support decision-making • Analysed large, ambiguous datasets on ageing water meters • Delivered recommendations on optimal replacement timing, leading to future cost-saving strategies

Bachelor Thesis collaboration on cognitive impairment prediction modelling • Worked with a team of professors from the Faculty of Psychology and Neuroscience • Developed prediction models for cognitive impairment diseases with 80% prediction accuracy. • Implemented Classification, Regression, and Forecasting models on Proteomics Dataset. • Combined feature selection techniques with optimized models, resulting in increased final accuracy.

Early AI engineering internship focused on computer vision. • Built and trained CNN-based models for image processing use cases • Developed automated pipelines for image scraping, cleaning, and labelling • Increased usable image data by ~150% and delivered a process adopted internally