Internship details

• Duration: 6 months

• Start date: ASAP

• Location: Toulouse

• Team: Perception

• Internship subject: Deep Learning on accumulated LiDAR PointClouds

• Compensation: 1000€ gross, tickets restaurant, CSE

Internship Context

EasyMile is advancing Level 4 autonomous technology, allowing our vehicles to operate fully driverless in defined environments by independently managing a lot of driving situations. Safe operation requires a high volume of environmental information, which is gathered and processed in real-time by a full range of on-board sensors like LiDAR, RADAR, and cameras. These sensors create a complete 360-degree environmental model, capturing infrastructure, moving adverse, that the driverless system uses to make safe progression decisions.

In the Perception Team, we use deep learning techniques to analyze this surrounding environment. Currently, our scene understanding stack runs online using few sensor scans; however, accumulating LiDAR point clouds along the vehicle’s trajectory, especially with newer sensors, allows us to represent a dense 3D scene that can significantly elevate our final performance.

This internship focuses on applying Deep Learning techniques to these large, dense PointClouds of our vehicle environment. You will explore recent work like the Superpoint Transformer [1] and the lightweight EZ-SP [2] to help us redefine our offline perception stack, specifically aiming to improve our auto-annotation process for static obstacles, enhance the automatic creation of HD Maps, and potentially develop a new format for semantic prior maps for our autonomous vehicles.

[1] Robert, D., Raguet, H., & Landrieu, L. (2023). Efficient 3D Semantic Segmentation with Superpoint Transformer.

[2] Geist L., Landrieu L, Robert D, (2025). EZ-SP: Fast and Lightweight Superpoint-Based 3D Segmentation

Missions / Responsabilities

Under the supervision of his tutor, the intern will be involved in:

• Format the easymile data for large pointcloud training

• Study the literature to find relevant architecture and network

• Train and evaluate such network on possible different use cases: static object detection and/or segmentation, HDMap generation

There is no typical profile at EasyMile, we all come from different backgrounds and that is what makes us strong! Don’t hesitate to apply if you are motivated and interested by innovative transportation and technologies.

We are looking for:

• Student in computer science or machine learning or robotics

• Skills: Deep Learning, Computer Vision, Python

• Soft skills: Team spirit, autonomy, and curiosity

• Language skills : English and French

• 30 minutes call with a recruitment team

• Meeting with the team, technical tests 

• One hour interview with the tutor and a recruitment team