Anne-Marie Tousch

Leading a team working on machine learning for geolocation data, turning raw, biased data points into actionable insights.

Anomaly detection (& potentially anything ML) for devops.

Skills: Data Pipelines · Python (Programming Language) · Predictive Analytics · Coding Standards · Technical Presentations · Statistical Data Analysis · Pandas (Software) · Data Science · Unstructured Data · Soft Skills · English

Having a full 3 years research experience in computer science. My PhD research was about using vocabulary structures (generically denoted as "semantic hierarchies") to help automatic image annotation. Basically, the idea is to fall back to a more generic description of the image when giving details is too difficult. I've used a database with cars, so it translates as, I can tell with more confidence that this car is a Peugeot, and less confidence that it is a 206 (and, with even less confidence, it is a 206-SW). Practically, I implemented a car recognition algorithm which followed an ontology on car models. I played a lot with SVMs during those 3 years (one-class SVMs, binary, multiclass, and more). I used LibSVM and its variants, VLFeat, and Imagine lab library (CertisLib, at the time). A fun part of my programming task was that I was working 2 days a week at IMAGINE, where i was running the code on Windows, and 3 days at ONERA, where i was running the same code on Linux. Cross-platform all along :-).

Development of a method that automatically extracts an object from an image in a single click, using geometrical and color properties (C++/Qt). Context: develop an easy selection interface for a robotic arm to be used by disabled people.