Developped AI models for real-time analysis of surgical videos. Leveraged Computer Vision models such as Dynamic Convolutional Neural Networks (CNNs) or Image Segmentation models such as Detectron2/SAM.
•
Built robust end-to-end data pipelines for ML models automation on thousands of hours of footage.
Research NLP internship focusing on non-labeled data.
•
Scraped (via Python: Selenium, BeautifulSoup and APIs) and used more than 10,000 working agreements from French government website. On the one hand, used those to understand the impact of a 2018 French reform on equality-at-work, and on the other hand to understand the impact of lockdown on the generalization of teleworking.
•
Worked with Datamatics for the labelling process, then employed supervised Machine Learning algorithms for exploiting the labelled data.
•
Also considered several other approaches such as NER (Named-entity recognition), Semi-Supervised learning, keyword searching, etc.