DuckDuckGoose / Delft (NL)Apply on site
As an intern, you will work very closely with the founders of DuckDuckGoose to improve the quality of the deepfake detection product. There are three (3) different tasks that we could use your help on:
1. Synthesized audio detection. Within deepfake detection, this is a relatively new research area because most deepfake detection models work with images or video. You will do this by researching the current state-of-the-art of synthesized audio detection, and you will apply the acquired knowledge to implement a prototype.
2. Neural network generalization capabilities. You will investigate the ability of deepfake detection models to generalize to new, unknown, deepfake generation methods and suggest improvements for these deepfake detection models. You will do this by researching the current state-of-the-art of out-of-distribution performance of neural networks and applying the knowledge gained to deepfake detection.
3. Neural network explainability. In addition to making a prediction whether image material is a deepfake or not, we at DuckDuckGoose also provide insight into what our algorithm bases this decision on: the so-called explainability module. Within this internship, you will research the current start-of-the-art of explainability of neural networks, come up with improvements where possible and find out how this can be applied to deepfake detection and our insight module.
During your internship, you will work on one of these tasks.
We provide deepfake detection products and services to the forensic industry. Our software is used by the Netherlands Forensic Institute and the National Police, among others. In addition, we want to grow into the digital identity verification sector in the short term.
* You are currently doing an MSc in computer science, data science, artificial intelligence or something similar.
* Experience with Python programming and demonstrable experience with the use of machine learning libraries such as PyTorch or Tensorflow. Experience in the field of computer vision is a plus.
* Good English communication skills.
* Looking for an internship of at least four months.
Conditions and benefits
* Impact: our product is used in various places to make a positive impact. You can contribute greatly to this.
* Working in a dynamic startup environment where you won't be bored for a day.
* A responsible position with a lot of freedom and flexible working.
* A workplace at our office in YES!Delft in a young and energetic team.
* Part-time or full-time in consultation.
* Internship compensation is negotiable.
Contact Joris at firstname.lastname@example.org. When contacting us, please indicate a preference for the 1st, 2nd or 3rd task.