machine learning

Freedom to Fumble: Learnings From Hackathon 2018

I spend all my time at work dealing with software. So going into this year’s hackathon, I was excited to get hands-on (and in over my head) with hardware.

I spend all my time at work dealing with software. So going into this year’s hackathon, I was excited to get hands-on (and in over my head) with hardware. Having never even touched a Raspberry Pi before, I was in for an interesting couple of days!

Our team was just two people—me and Shivani Thakar. We set out for a small, achievable project broken down into phases.

  • Our first step was to set up a Raspberry Pi 3 Model B and Camera Module v2

  • Our second step would be to actually capture and store images with the camera

  • Our third step was to submit those images to Microsoft Cognitive Services for authentication, identification, and recognition

Day one

Day one was all about the hardware: mostly, trials and tribulations—heavy on the trials.

Hands-on work with the Raspberry Pi introduced us to a new world of troubleshooting beyond just code—wrestling with USB-C adapters, Wi-Fi connectivity, and voltages.

Multiple times we seemed to work our way to a dead end and would have to backtrack and redo our steps. We made a special effort to document all our troubleshooting and resources so that we were able to approach the next iteration on sure footing.

It was essential to have a teammate and to bounce ideas off each other. Shivani and I had not really had the opportunity to work together directly before. I wouldn’t have learned so much without her. And other Toads were excited to lend our team a hand, too. The freedom to try, fail, and try again—with a little help from our friends—created an ideal environment for innovation.

Day two

On day two, when we finally got our first photo processed, it was so exciting—all our work came together!

Exploring the potential of Microsoft Azure Cognitive Services was fascinating. Their offerings do the heavy lifting for the back end of machine learning, which opens up so many possibilities for applying image recognition to real-world scenarios.


We got plenty of accurate results, but also had a lot of laughs with the silly ones. My favorite machine-learning miss was seeing a picture of a Teenage Mutant Ninja Turtle figurine identified as “a woman standing in front of a mirror posing for the camera.”

The algorithm had a pretty high confidence score about this wrongheaded result. And despite my own lack of hardware experience, I wound up with more confidence too at the end of the day. I might not have become a Raspberry Pi expert, but next time I get my hands on one, I can certainly fumble a lot better.


Our Internet of Things (IoT) project brought together a few core technologies—a Raspberry Pi, cloud computing, and image recognition software. Our experiments gave us lots of ideas for how these fundamentals could be applied to the real world, like:

  • a visitor greeting system that would identify employees and guests entering the office

  • a portable handheld device that could identify objects and emotions in the real world

  • a zoo or arboretum exhibit that identifies the flora and fauna in real time

This hackathon granted me familiarity with new tools that I can draw on for future Metal Toad projects—especially as IoT becomes more and more important for our clients. My increased understanding of the hardware opens up a lot of insights for connecting the real world to the digital.

But the biggest takeaway from the hackathon, for me, was strengthening the bonds that help our team innovate and solve problems together. The freedom to experiment, fail, and fumble together gave us a chance to flex our creative-thinking muscles and strengthen our collaborative spirit.

Similar posts

Get notified on new marketing insights

Be the first to know about new B2B SaaS Marketing insights to build or refine your marketing function with the tools and knowledge of today’s industry.