Hi, I’m Axel Mendoza. A data scientist based in Paris. I’m deeply passionated about Computer Vision and meditation. My greatest satisfaction is to build AI products that improve people’s quality of life.
DOWNLOAD MY CV
Designed person tracking and re-identification systems of an autonomous grocery store. Enchanced product attribution using Pose Estimation. Created a semi-automatic annotation pipeline to generate data for Deep Learning using Airflow and GCP. Managed and trained a team of 4 annotators on bounding box and pose video annotation.
Improved physician diagnosis of heart disease by creating a coronary calcium detector trained on CT scans using Unet with Pytorch. Improved patient’s disease evaluation even further by classifying calcium in high and low risk arteries. Optimized model complexity to fit hospital needs by designing a faster approach using ResNet3D. Enhanced detection of mitral valve regurgitation by creating a blood flow dealiasing model using Unet trained on 3D color doppler data.
Improved security of power-plants by designing a multi-camera vehicle re-identification and tracking system using Keras and TensorFlow. Implemented 2018 state-of-the-art solution and improved mean average precision by 6% by adapting a pedestrian re-id paper to vehicle tracking. Collaborated with the best researchers in the field after being invited to ECCV 2018.
Implemented an automatic testing platform for SAP products.
Top 1 computer engineering school in France.
Artificial intelligence major.
Exchange student for 6 months
0
WORK HOURS
0
BLOG POSTS
0
POSTS LIKES
0
COFFEE AT WORK
State-of-the-art implementation of a multi-camera vehicle re-identification. The main challenge of this project was to build a model able to extract features of cars in real-time to recognize the exact same car in different cameras. The model, trained on triplet loss, is capable to recognize fine-grained details: broken wheel, objects on the board, or stickers to differentiate two cars with the same model and color in a 360° fashion. I implemented the state-of-the-art of 2018 and adapted a pedestrian re-identification paper to improve the solution.
READ MORE
Coronary calcium scan is a test that measures the amount of calcium in the walls of the heart’s arteries. A patient with a high quantity of calcium in his arteries will inevitably die without the proper treatment. The goal of this project was to build a coronary calcium classifier to predict the severity of the patients heart disease. The model was able to predict for every calcium point in which artery it was located as it is very important to know if the calcium is in the main or in minor arteries. Trained the model on 3D X-ray data using a ResNet3D.
READ MORE
I designed an autonomous driving remote car by assembling a Raspberry on top of a remote-controlled car embedded with a Convolutional Neural Network. The model was trained on some data captured while driving the car using an xbox controller. This project was held during my last year at EPITA as a graduation project and was selected to be shown at my university open house. I took the project again a few months later and built a more sophisticated car along with a data generation module. I've won the RobotCars Winter 2018 championship and got ranked 3st at IronCar Summer 2018 tournament.
READ MORE VIDEO
From scratch implementation in Pytorch of the most used machine learning algorithms. Each method is clearly explained using latex formulas and fully commented code. Every post can be found in my blog as well as on my Github. I implemented the following algorithms: support vector machine, decision tree, random forest, adaboost, k-means, gaussian mixture model, naive bayes classifier, polynomial regression, k-nearest neighbors, logistic regression, and linear regression.
READ MORE
CUDA C++ is a GPU programming language that uses a grid based approach to parallelisation. Each GPU unit has a coordinate in the grid and computes a sub part of the process that allows a massive computing speed up. I implemented blurring, de-noising and edge detection algorithms from scratch using only exclusively the CUDA C++ standard library.
READ MORE
I had the pleasure of working with Axel during his apprenticeship at Siemens Healthineers as his research supervisor. Axel is an independent and quick learner, who was able to rapidly learn new concepts he was working on. He maintained good communication with the team, successfully completing the tasks assigned to him on multiple projects. Overall, it was a pleasure to work with him.