Dr.-Ing. Oliver De Candido

Senior Research Engineer @ neurocat GmbH.

After growing up in Hong Kong and completing my GCE A-Levels at the German Swiss International School, I came to the Technical University of Munich (TUM) to pursue higher education. I obtained a B. Sc. and an M. Sc. in Electrical Engineering and Information Technology in 2014 and 2017, respectively. From 2018 till 2023, I was a research assistant at the Professur für Methoden der Signalverarbeitung (MSV) at TUM. I was also an external doctoral student at AUDI AG from 2018 to 2021. I graduated with the Dr.-Ing. degree (summa cum laude) in Electrical and Computer Engineering in 2023. My research focuses on validation safety arguments for machine learning based highly automated driving functions. Since July 2023, I have been working at neurocat GmbH as a senior research engineer and team lead.

If you want to chat or know of exciting opportunities related to AI/ML safety, please do not hesitate to reach out!

Education

  • Dr.-Ing. in Electrical Engineering and Information Technology, Technical University of Munich, Germany (Apr. 2018 – Oct. 2023)
  • M. Sc. in Electrical Engineering and Information Technology, Technical University of Munich, Germany (Oct. 2014 – Jun. 2017)
  • B. Sc. in Electrical Engineering and Information Technology, Technical University of Munich, Germany (Oct. 2011 – Oct. 2014)
  • GCE A-Levels, German Swiss International School, Hong Kong, Hong Kong (Jun. 2011)

Research

My main research interests lie in Machine Learning (ML) and Artificial Intelligence (AI), specifically, working on methods to support a safety argument for safety-critical applications. Previously, I researched signal processing techniques for Multiple-Input Multiple-Output (MIMO) communication systems.

A list of my publications can be found here: Google Scholar or mediaTUM.
You can read my dissertation here: TUM Diss.
The code from my research can be found on: GitHub.

Validating Machine Learning-based Highly Automated Driving Functions by Diversity

The main focus of my doctoral project was researching ML safety methods used to build validation safety arguments of ML-based safety-critical driving functions. I built safety arguments by validating various aspects of deep neural networks. In my research, I focused on the following methods: representation learning, self- and semi-supervised learning, clustering methods, interpretability methods, and the distributional shifts between public highway driving datasets. I primarily worked with time-series data on object lists.

© A. Rosebrock, CC BY-SA 4.0

Utilising Image Augmentations at neurocat

At neurocat, I worked with my team on using realistic image augmentations generated by aidkit to: (i) build safety arguments of perception models; or (ii) fine-tune perception models with augmentations to improve performance in specific operational design domains. We primarily worked with open-source object detection and semantic segmentation models and datasets, e.g., Faster-RCNN, RetinaNet, DeepLabV3+, or the ZOD dataset. We built robust reporting methods which could be used in safety arguments and found training strategies to improve models' performance in specific cases.

Machine Learning/Artificial Intelligence Safety

Besides my doctoral research topic, I am interested in ML/AI research. I have worked on various projects related to ML safety, e.g., using reinforcement learning in automated driving or detecting adversarial attacks in modern ML models. I am also interested in AI safety technical research.

Teaching

I was also a teaching assistant at MSV, working with two graduate level courses:

I completed both the foundation level certificate and the advanced level certificate for Teaching in Higher Education at Bavarian Universities. I took courses ranging from the theories of teaching and learning, through to advising and counselling students.

Whilst at MSV I supervised: 8 Master's Theses, 6 Bachelor's Theses, 6 Research Internships, and 7 Seminar Papers.

Facilitating

In my free time, I have found it rewarding and engaging to help facilitate (and participate in) various online courses related to AI/ML safety. Below are the courses I've facilitated with a link to the course for more information!

Contact

If you want to contact me about my research, teaching or other experience, reach out via LinkedIn!