Valley DevFest 2019

Sep 25, 2019

Valley DevFest 2019 was a great conference, in my opinion. I broke my promise of not giving a talk when I’m responsible for the website. I also worked on an Augmented Reality companion app (see my earlier blog post about it) until 3 AM, and I had to fill in some session spots. This year, we didn’t have any Googler presenters, so we had to produce more talks on our own.

I had an individual talk about Git workflows, which was recorded by GMAC. My main goal was to give attendees a deep understanding of how a simple tool like diff plays the fundamental building-block role in large ecosystems such as GitHub, GitLab, Bitbucket, and the like. I started with a brief history of source control systems so we could see where we are compared to the past. The secondary goal of the talk was to show standard branching and workflow conventions and tools like GitFlow (see also this excellent cheat sheet) and Git Town. These tools simplify complex multi-stage git command series so you can follow branching and merging conventions much more easily. In turn, these also boost collaborative work and software quality. The pinnacle of this is server-side merging with Git Town (yes: the merge actually happens on GitHub, and if something fails, the PR is rejected). See more in the slide deck of the talk.

I filled in two more presentation slots, though: I organized two sessions of Machine Learning lightning talks. Those two sessions featured four student projects each, presented by the students of the Fresno Data Science meetup:

  1. Machine Learning Lightning Talks - pt 1
    • Lourdes Lopez: Predicting Animal Shelter Adoption Outcomes
    • Paul Settle: Our Children: Predicting a Child’s Ability to Pass the CAASPP
    • Martin Kunz: Predicting Work Flow
    • Beth Weinman, Csaba Toth, Grace Liang, Elizabeth Gaw: Can We See Climate Change in Fresno?
  2. Machine Learning Lightning Talks - pt 2
    • Federico Ulloa Rios: Bearing Fault Detection through Convolutional Neural Networks
    • Csaba Toth: Unconventional analysis of the Ridgecrest earthquakes
    • Jorge Ovando: Predicting Song Popularity using Spotify Track Features
    • Dale Herzog: Applied Machine Learning in Oncology

I’m really proud of all of the students. I know several of them will be successful and will pursue data science further. Lourdes Lopez, for example, already presented Intro to Python Libraries for Data Science at the Tracy Developer Meetup.

As you can see, not only did I organize the lightning talks, but I also had a part in a presentation in each. Talking about climate change in Fresno, we basically started to scratch the surface, but we were able to show evidence of temperature rise by looking at the data with the H-index technique: we counted how many days the temperature was above a critical level each year. See slide deck and iPython notebooks in this directory.

As for the Ridgecrest earthquake series, I used PCA (Principal Component Analysis) to find the two new perpendicular fault lines, which emerged nicely from the data. By inputting the lat/long coordinates, PCA found the fault lines, in my opinion. Besides that, I wanted to see how one month of earthquakes unfolded in time and space. The web demo illustrates that in an extreme fast-forward manner, but the size and color of the dots and also the volume of the pop sounds reflect the magnitude of the shakes. Materials:

In hindsight, I pretty much overworked myself (yet again) with all of this; I could have performed better at the Git talk at the end of the day if I hadn’t shown my car at the electric car week the night before, if I hadn’t worked on the Augmented Reality Map app until 3 AM, and if I hadn’t had the two other lightning talk sets. But I’m really happy that I pushed through; I was able to elevate people at the lightning talks and still fare decently while doing approximately 3-4 people’s workload. I also know that other organizers put in more than 100% on their parts to make this event as great as possible.

Our next conference is the IWDC (International Women’s Day Conference) in the spring of 2020.

Comments loading...