NAVYA

My responsibility was to build a validation database based on a month’s worth of observed ground truths of one autonomous vehicle’s test rides on a site in Paris, and from there produce a testing script that evaluates post-processing detection and classification algorithm. These algorithms classify log data received from the autonomous vehicle’s journeys and predict various driving situations. My work directly contributes to the iterative development pipeline of said algorithms, providing a performance index to base improvements on. Afterwards, I had the time and opportunity to directly develop and improve one of the algorithms: applying optimisation methods, exploring the design space, and selection based on performance and interpretability. Ultimately, these improvements are to create an autonomous data analysis pipeline that identifies and mitigates and abnormal situations, ensuring that the company’s products are fully autonomous and safe during service.