Tomas Moitinho de Almeida Andrade Madeira

» Developed a multi-modal RAG system with 5+ data sources, including web search and in-house content farm. Implemented agentic reflection for continuous content relevance, resulting in a 20% accuracy improvement. » Utilised GCP multimodal embeddings and Vertex AI to enhance matching between LLM output and relevant images, achieving 95% precision. » Led company-wide AI strategy, outlining product portfolio and 12-month road-map.

» Led team and managed project of OCR pipeline for processing 1M+ documents in 30+ languages using AWS & Azure Cognitive Services. » Deployed NER micro-service for real-time data extraction, saving 100+ hours annually. Implemented modular design for cross-departmental re-usability. » Developed Topic Modelling and Search Engine tools for R&D, enabling visualisation of 10k+ papers for research discovery. » Supervised NLQ master thesis, providing mentorship for junior colleagues.

» Published peer-reviewed research on NLP, introducing a novel classification framework for linking ML to aviation safety. » Trained embedding model from scratch with domain-specific corpus, benchmarked supervised (SVM) and semi-supervised (LS) algorithms, and tuned hyper-parameters with Bayesian Optimisation. » Influenced further academic studies by spearheading advancements in reducing aviation incidents through advanced human factors analysis.