Approximately two weeks prior to OSCON I met with cofounder Vinay Rao regarding ways I could potentially become involved with RocketML. Both of us had been involved with the Portland Data Science community for quite some time and after a lengthy discussion we honed in the idea of Hackathon to both demonstrate and promote RocketML. Furthermore machine learning is a tedious process, any approach that helps improve data scientists productivity is good for the community and industry in general. That was a brainstorming session without a definitive action plan!

A week later, Vinay contacted me, “How do you feel about trying to do the hackathon idea at OSCON as a pilot?”

I said, “Sure, sounds like a great idea!”

“OSCON is next week..” he said.

“Oh… OK,” I said feeling a bit of hesitation. I had never been involved with a hackathon before, not to mention running one, so having to help and organize one so quickly gave me pause.

Vinay concluded, “Great! Let’s meet this coming Saturday and get this thing done. Until then I would recommend learning Tensorflow.”

“Well I’ve used Keras before and have always wanted to learn Tensorflow, now’s a good time as any.” I said. There I go…. I felt like I jumped off the cliff!

There was no time for buyer’s remorse. I started an online tutorial to help me to get familiar with Tensorflow syntax because I had already heard the documentation for Tensorflow was sparse and difficult to follow at best.

Following day, I spoke with Santi Adavani (cofounder, CTO of RocketML) and he confirmed the need for me to focus on Tensorflow, and additionally asked me to focus also writing OpenCV tutorials as most likely we would be using object detection on images and videos. Once again I had concerns about how this was going to work, since I had never worked with video or image data before, but Santi seemed confident so I trusted his judgement. Later while browsing through online tutorials I noted down relevant materials and started building a Jupyter Notebook that I might want to use for guiding others with using OpenCV; this proved invaluable in terms of time management.

Lesson #1 : Get used to Jupyter notebook fast. It is the lifesaver for any wanna-be data scientist!

Following day, I spoke with Santi Adavani (cofounder, CTO of RocketML) and he confirmed the need for me to focus on Tensorflow, and additionally asked me to focus also writing OpenCV tutorials as most likely we would be using object detection on images and videos. Once again I had concerns about how this was going to work, since I had never worked with video or image data before, but Santi seemed confident so I trusted his judgement. Later while browsing through online tutorials I noted down relevant materials and started building a Jupyter Notebook that I might want to use for guiding others with using OpenCV; this proved invaluable in terms of time management.