My Pumas-AI Perspective
First of all, I loved the experience of the internship, it is not my first time working fully remote with people I have never met, and obviously that is not always easy at times. However, I believe due to the nature of the people at Pumas, this experience was as good as it could have been. This is especially evident in the biweekly calls we have with everyone interested in the Bayesian side of things that are incredibly insightful, and honestly some of the best and most enjoyable discussions I have been a part of since starting my PhD. I look forward to hopefully still dropping in on those to talk about the work I have done and the plans we have to continue on with my capacity as a contributor to Turing.jl now, or future collaborations that may pop up.
On that, I definitely have achieved my goals in terms of not only building something useful to Pumas but also getting to engage with a community I already knew I was very aligned with in the Turing PPL team and larger Julia community. Mohamed has been an amazing mentor and friendly face throughout. Everyone was super understanding and kind, whilst still I think letting me properly get to grips with the work and build a package I am proud of and will continue to contribute to for some time to come. With the ever-growing popularity of the Turing PPL, the package itself should hopefully be seen as a great contribution to the Julia community from Pumas.
The support in terms of pair programming, having my questions answered directly and clearly, working through problems both theoretical and in code etc. was all great, and I would definitely recommend the experience to anyone that wants to be truly working at the nexus of academia and industry. The experience did feel very similar to other collaborations I have had during my PhD which for me was a huge plus, as it felt like my time was not unnecessarily wrapped up in meetings or other activities due to the clarity and purpose of the company leading to a very natural and organic research-focused-yet-pragmatic working approach. I feel like a much more competent Julia programmer (and programmer in general) through the code review and time spent on calls with Mohamed and others; these skills clearly transfer into all of my work so I am grateful to have been given the time and patience to fully build on my existing skills in this way. On the theoretical side too, I definitely feel a lot more confident in this methodological area and feel lucky to have been able to ask questions of Hong and others in the field that I respect.