CV

Performance Development Engineer

Audi Formula One project (Sauber Motorsport AG)

Webapp software developer. F1 Performance Engineer for Nico Hülkenberg specialised in live driving lines for Practice, Qualifying, and Race. Data acquisition of telemetry for competitors during live sessions

• Software Development
• Typescript
• Svelte
• Machine Learning
• Driving Lines
• Data acquisition
• Data management

Data Analysis Intern

Audi Formula One project (Sauber Motorsport AG)

Automatic extraction of trajectory information from on-board F1 footage using Deep networks and Visual SLAM

• Data Analysis
• Machine Learning
• SLAM
• Statistics
• Visual Odometry
• State Estimation
• Pose Estimation
• Jira
• GIT
• Scrum Master

Student Research Assistant

Medical Image Analysis Lab (CHUV, Lausanne)

Trained and deployed deep models for the detection and classification of lesions in brain MRI images. Developed a software to make the network easily accessible to clinicians with no prior coding experience, using docker (available in projects section)

• Data Analysis
• Machine Learning
• 3D images
• Software development
• Docker
• Medical Imaging
• Statistical Analysis

Head of perception, driverless division

Supervised and led a team to develop the vision/perception pipeline for a self-driving racing car with important constraints: the code should be robust to potential sensor failures, accurate and run in real-time on an embedded computer (Nvidia Jetson Orin).

• Computer Vision
• Machine Learning
• LiDAR
• Monocular Camera
• Real-time
• Sensor fusion
• Sensor Calibration
• DBScan
• Nvidia Jetson
• TensorRT
• Ransac
• PnP
• Testing and Debugging
• Team Management

Master's

Ecole Polytechnique Fédérale de Lausanne

Electrical Engineering, specialization: signal, image and video porcessing

• Deep Learning
• Applied Data Analysis
• Computer Vision
• Modern NLP
• Network Machine Learning
• Lab in Android development
• Image Analysis and Pattern Recognition

Bachelor

Ecole Polytechnique Fédérale de Lausanne

Electrical and Electronics Engineering

• Signal Processing
• Power Electronics
• Energy Distribution
• Signal and Systems
• Control Systems
• Electromagnetism

Streamline RimNet: Tools for Automatic Classification of Paramagnetic Rim Lesions in MRI of Multiple Sclerosis

MIAL, CHUV

Abstract to promote the software. It deploys the RimNet deep network to clinicians, where they would only need to click on a potential lesion to extract a patch, run the network in realtime using docker, and receive instantaneously a feedback

Automatic Extraction of Yaw Rate and Velocity from on-board Formula One footage

Extract trajectory information (Yaw Rate and velocity) from monocular on-board Formula One footage.

• Visual SLAM / Visual Odometry for State and Pose estimation in C++
• Trained / Deployed deep networks using Python and Pytorch.
• Webapps development using Gradio for data visualization
• Statistical analysis on the results to evaluate the robustness of the experimented methods

EPFL Racing Team Driverless perception

Supervised and led a team to develop the vision/perception pipeline for a self-driving racing car with important constraints: the code should be robust to potential sensor failures, accurate and run in real-time on an embedded computer (Nvidia Jetson Orin)

• Developed a real-time object detection and distance estimation algorithms, using a monocular camera and a LiDAR.
• Performed object and keypoints detection, as well as PnP for distance estimation using just a monocular camera
• Performed Ground removal using Ransac, DBscan clustering, ego-motion correction on the LiDAR pointcloud
• Performed sensor calibration (obtain intrinsic and extrinic parameters) and sensor fusion using computer vision projections (3D-2D) to obtain better results
• Integrated robust and realtime algorithms to the main pipeline with ROS2
• Deployed, tested and debugged the algorithms for realtime use on a Nvidia Jetson

Streamline RimNet: Tools for Automatic Classification of Paramagnetic Rim Lesions in MRI of Multiple Sclerosis

Improve the current network for the classification of brain lesions, and find a good and easy way to automatically deploy it to clinicians without prior coding knowledge required. Resulted in an abstract publication at ECTRIMS 2023

• Trained deep networks for the detection and classification of Multiple-Sclerosis lesions in brain MRI images with Pytorch
• Performed data pre-processing and normalization using biomedical imaging frameworks such as Monai
• Developed a software tool to deploy the model to clinicians using docker. Software extracts a patch around a click, feeds it to the network and provide live feedback to the user
• Published an abstract at the ECTRIMS 2023 conference, regarding the robustness of the deep model with respect to the center patch selection

Android application: GymRat

Free, opensource and fully local android application written in kotlin to track exercises at the gym, as well as the bodyweight, calories and protein

• Kotlin
• Fitness

Android application: MyMeds

Free, opensource and fully local android application written in kotlin compose to track available medications at home, expiry date and category

• Kotlin
• Kotlin Compose
• Health

ApiZoom – deep learning to quantify the Varroa parasite in honey bee hive images

Bachelor thesis with Prof. Jean-Philippe Thiran. Main task: automatically detect toxic bee-killer parasites on bee-hive images.

• Computer Vision
• Metric analysis: Precision, Recall, F1, mAP, AUC ...
• Successfully trained a YOLOv5 network for the automatic detection of the toxic varroa mites