Vivien Mallet

I am a data scientist and ML engineer with a background in statistical modeling, machine learning, and scientific computing. I am seeking an AI position, with a particular interest in LLM-based systems and modern AI applications.

I am a former researcher in statistical modeling at INRIA, France's national institute for research in digital sciences and technology. Since 2020, I have worked remotely as a freelance data scientist and ML engineer.

My latest professional position was Machine Learning Lead at Jane's Weather for over two years. I developed a deep-learning-based weather forecasting system across Australia, ingesting over a million observations per day and serving forecasts for thousands of locations. The role ended when the company faced severe financial difficulties. I have also developed data-intensive services in other domains, including financial auditing with accounting datasets of up to 200 million records.

My core background is in numerical machine learning, uncertainty quantification, forecasting, and data-driven modeling. More recently, I have extended this experience toward LLM systems and modern NLP-oriented AI engineering through substantial personal work. In particular, I developed two public projects: the multilingual FastAPI application "Word Teacher" (https://git.vivienmallet.net/mallet/word_teacher) for LLM-generated vocabulary quizzes, with offline fine-tuning and evaluation, and online tracing; and "News Curator" (https://git.vivienmallet.net/mallet/news_curator), a generator of daily news briefs relying on a vector database, evaluated LLM, online evaluation with masked LM, and cross-encoder-based deduplication.

I would like to continue in this direction in a position involving generative AI, either fully or as part of a broader AI role. I believe I can bring to a team a combination of strong mathematical and machine-learning depth, and the ability to design innovative, rigorous and operational AI systems.