My expectations and motivation for the PFN internship
The reason I applied was because I thought I would enjoy working with very talented engineers and researchers and having discussions with them.

About the internship atmosphere
The internship was online due to COVID-19. We had a daily meeting in the late afternoon, so I worked every day trying to make some progress before the meeting.
PFN employees kindly arranged what they call “shuffle lunches” for interns. In shuffle lunches, I had a chance to chat with different interns every time. It was nice to get to know and make friends with other interns.
I worked from morning to evening, and at night I spent time playing games with other interns. I really enjoyed the time of my internship.

An unforgettable experience during my internship
At the first mid-term presentation, other interns seemed to be making much more progress than I was – I was like, “Oh no…” That put me under a lot of pressure. 
Then I decided to forget about the first two weeks and completely shifted my focus to another research project in the second two weeks.
At the end of the day, I was able to produce an interesting result and presented it with confidence at the final presentation.

What I gained from the PFN internship
The result of my internship research was submitted to an international conference and was accepted (See below). It even received the best paper award.
I still keep in touch with other interns and sometimes hang out with them.

My internship theme and year
I participated in the PFN internship in the summer of 2022. I was a first-year student in a master’s program. I chose the theme JE14 “Sensory data compression with deep learning”

My expectations and motivation for the PFN internship
I applied for the internship program, expecting to gain knowledge in a field that I had not studied before and to improve my technical skills further. I also found it significant that I would be involved in the development of technology for products made by a leading AI company.

About the internship atmosphere
In my case, I worked mostly online and mainly with my mentors and people related to the project. They were very friendly and helpful, and took the time to explain technical details and answer my questions.
On days when I worked in the office, I had lunch with other interns. We also had video conversations via an online app called Gather when working from home.

A day in the life of an intern
Introduction:
In my case, the data used for my experiments were too large (more than 1,000 pieces of 150MB data each) to do the experiments by myself, so I had to plan and coordinate with my mentors properly. Usually, however, you should be able to work directly with computing clusters by yourself, so you do not need to plan that much. Please read the following with this point in mind.
9:00〜11:00: Review the results of my experiments and check what ideas and tweaks have worked well. Based on them, I do some research on a better way to run the experiments and implement the idea.
11:00〜11:30: Daily meeting with mentors. Because I only have a short amount of time to give a summary of my experiments and schedule, I make sure I know exactly what I am going to say, so that the meeting goes smoothly. I also try to seek their advice and suggestions during the meeting, which is very helpful.
11:30〜12:00: Based on the outcome of the meeting, I clarify what should be implemented and prepare for the implementation. If I am given a reference paper, I read the outline and the sections that seem to help the implementation.
12:00〜13:00: If I am in the office, I have lunch with my mentors and other interns. We buy lunch boxes or sometimes eat at a nearby restaurant if it is not crowded. 
13:00〜15:00: Implementation work. In many cases, things become easier afterwards when I consider modularity and implement with care so as not to adversely affect past experiments.
15:00〜16:00: Test the implemented code with a small amount of data to see if it works correctly. If it works well, upload it to GitHub and make it available to the server.
16:00〜17:00: Describe in a file what parameters to use in the experiment, and ask my mentors to run. If an error occurs, I ask them to try it again after fixing the error. 
17:00〜18:00: Make preparations to implement what I could not implement today, or read a reference paper.
18:00: Report my progress to my mentors and finish for the day.
Interns also have several opportunities to give presentations, so I prepare slides and practice for the presentation in my spare time.

An enjoyable/interesting experience in my internship
What I found interesting about the PFN internship was that I could use supercomputers as much as I wanted. Of course, you can’t use them for private use, but you can use many computing resources if it is related to your internship project. So, it is very fun to be able to do many things (e.g., running in huge batches). To run experiments, we use kubernetes to send jobs, so it’s easy to use and my research progressed very well!

Challenging but rewarding task
Since the task of compression was completely different from my research in graduate school, it was challenging to learn the prerequisite knowledge, and then to actually code and implement it. At the same time, it was fun to experience new technology and broaden my knowledge, and I felt a great sense of accomplishment when my implementation was accepted!

Beyond my imagination
The entire research system was different from what I had expected in a good way: it was very sophisticated and easy to use without much difficult configuration. I got used to it within a day or two. I had never used Docker or kubernetes before, but I was able to do almost everything by myself without having to ask. I think the system was very helpful that even someone like me who had never used these technologies could handle them without worrying.
Also, every one of the PFN employees I worked with was really skillful, and their amount of knowledge and implementation skills were much higher than I had imagined. By working with these professionals, I was able to not only acquire skills and knowledge, but also get a sense of what it is like to be an employee at a company that leads the world with its technology.

What I gained from the PFN internship
I was able to learn new skills in basic technologies such as Docker and kubernetes, which I had never used before, and also broaden my knowledge by implementing a thesis in an entirely unfamiliar field.
In addition to that, I was able to learn practical development techniques such as writing code that is easy to read for others and creating demos while imagining how it would actually work as a product.

Writing user-friendly code
Although my university research and internship work are not closely related, I have learned to always think about how to write programs that are easy for other programmers to understand. For example, now I try to simplify my code and avoid giving multiple roles to a single function or class rather than coding however easy for me.

Published a blog article and a research paper
I presented the accomplishments of my internship project at the 2023 annual meeting of the Institute of Electrical Engineers of Japan. I also plan to make a presentation at a conference held by the Meteorological Society of Japan.

PFN internship is highly recommended!
I recommend the PFN internship especially for those who like to develop and work with deep learning-related technologies! If you are interested in 3D Scan, CuPy, Optuna, or anything else that PFN creates, PFN internship would be a great opportunity for you!