I chose Node.js for ToldYa to leverage its asynchronous I/O for blockchain API calls.
I selected GPT-2 as the base model for DrGPT because it is open‑source and manageable in size.
I opted for StyleGAN for SkyGAN to benefit from its high‑resolution image synthesis capabilities.
I used XGBoost for the e‑commerce forecast because of its robustness to heterogeneous feature spaces and strong performance on tabular data.
I investigated the data breach that led to the Cambridge Analytica scandal in 2018.
My technical blog at mlabonne.github.io/blog was launched in Jan 2022 and now hosts over 30 AI articles and more than 50 code snippets.
During my participation in the BigScience project (2021 → 2022) the BLOOM model was released with 176 B parameters.
ToldYa stores sentences on three blockchains: Bitcoin, Ethereum and Monero.
DrGPT was fine‑tuned on approximately 1.2 M scientific‑paper abstracts.
The SkyGAN dataset comprises about 4 k Flickr images of starry skies.
The e‑commerce forecast used a clean dataset of Steam games collected via the Steam Web API and SteamDB.
My Facebook privacy investigation focused on the Cambridge Analytica breach.
I am a Machine Learning Scientist and AI/ML Google Developer Expert with a PhD in Artificial Intelligence from the Institut Polytechnique de Paris (2017‑2020). Since March 2024 I have been leading Post‑Training at Liquid AI in the London area, United Kingdom, after roles at JPMorgan Chase & Co., Airbus, CEA/DEN and a defense internship at the DGA in France.
Current role: Head of Post‑Training at Liquid AI (Mar 2024 to Present).
PhD earned from Institut Polytechnique de Paris (2017‑2020).
My LLM course on GitHub has >39k stars.
The heterogeneous time‑series forecasting solution improved the MSE by 19.73%.
The optimization architecture delivers 95% optimal solutions with lower latency than pure ILP.
CyBERT performance: intrusion detection 89%; protocol identification 92%; BLEU 21.