We are about half-way through our project and, unfortunately, this means that our intern Bernhard has completed his internship, and is leaving LiviaAI. Bernhard’s skills, creativity and commitment have been crucial to our effort. His uncomplicated, down-to-earth style, and his ability to explain complex things with ease and clarity have made working with him a pleasure. (Trust me, I have learned way more from him than vice versa!) All of the cool stuff showcased on our blog - collection metadata sentence embeddings or the automated triplet selection - are his work! Naturally, we are sad to let him go. But the good news is: there’s a ton of stuff that Bernhard has created, which we haven’t shared with you so far. The best is yet to come! But first, before letting Bernhard off the hook & back to university, we have asked him to write about his internship experience, and tell us about what LiviaAI was like for him!
My name is Bernhard Franzl and I am currently finishing my Bachelor’s degree in Artificial Intelligence at the Johannes Kepler University in Linz. I am currently writing my thesis, and I am about to start my Master’s in the upcoming winter semester. In the six semesters I have studied so far, I got to know a lot of different aspects of artificial intelligence. However, many topics are only covered briefly, and with little reference to practical application. Exercise courses, which focus on practice and implementation, are only held for selected important aspects of Machine Learning. I assume the reason is that the university simply does not have the resources to go into more detail. Moreover, I think the tasks - and especially datasets - in the exercises are a bit too simplified and with little practical relevance. Therefore, I felt the need to work on something more realistic, and I began searching for an internship.
I started working on the LiviaAI project in April 2022, and was instantly excited about it. My initial challenge was to find out how I could apply what I had learned at university to the tasks of LiviaAI. However, I quickly realised that my practical university knowledge would fit perfectly, and that LiviaAI was in fact a great opportunity to test my knowledge - and get some feedback on the way!
I found it refreshing to work with real-world cultural data. It was fascinating to get a look inside the museum collections. I think that the connection between culture and data science is what made this internship so interesting to me. In particular, I also enjoyed the opportunity to work with visual arts data, because I am particularly interested in computer vision and applications that involve image data.
I think that, most of all, my internship has improved my programming skills. Working constantly on the project’s many small programming tasks made it a routine. Furthermore, it was interesting to work with PyPI (the Python package management repository) to publish the main LiviaAI code as a Python package. This was a great opportunity to learn the basics about Python package maintenance.
Even though I feel I was well prepared for the project, I still had to do a lot of research. This was admittedly sometimes boring. But I learned about a lot of new algorithms, models and concepts. For me the most fascinating ones were Sentence Transformers, because they are just extremely useful. At the university, we were only briefly introduced to the underlying mechanism. Doing more research and applying them in practice turned out to be very interesting, and I gained a lot of new insights.
All in all, I am very satisfied with my internship at the Austrian Institute of Technology. It fulfilled all my expectations, and gave me the opportunity to test my knowledge on a challenging real-world problem. I am happy that I had the chance to be a part of LiviaAI.