Master thesis: Core Machine Learning

As part of an initiative to build a re-usable framework for machine learning at Volvo, we are conducting three inter-related 30p thesis projects for 2 students each.
There is a periodic need to modify the design of on-board embedded software. This means that code that follows one set of design principles need to be refactored to fit a new set of design principles. This thesis will study the refactoring process and will also analyze if machine learning can help automate or quicken any step in the process.
While training any model that learns vehicle behavior, the model also needs to be provided a fairly detailed description of the vehicle. Vehicles are usually described in the form of multi-variate discrete features which must also be embedded into the predictive model. This thesis will explore how to train such an embedding and also verify whether the concepts it learns aligns with expert knowledge.
Methods to estimate the speed at which the ego-vehicle is moving is usually error prone. Therefore, in order to increase the integrity of the estimation process, several methods are used in parallel to estimate the speed and the estimations from all these sources are fused. As an extra source of estimating speed, this thesis will focus on using camera images of the vehicle surroundings to estimate its speed.

Mandatory requirements:
Masters in Data Science, Computer Science, Software Engineering or similar
Skilled in programming (preferably Python)
Skilled in working with a Linux environment



Desired requirements:
Skilled in deep neural network modeling and training using PyTorch, TensorFlow or similar frameworks
Practical experience working with machine learning and/or deep learning and training CNNs, Transformers, etc.



General requirements:
An inquisitive mind
Ability to work closely with a team, while also being ready to take initiative
Flexibility in working with tasks in different application areas

For further information regarding these theses, you are welcome to contact Daniel Karlsson at 073-9029949.