Star Cloud

Star Cloud is a prototype that was the result of a collaboration with a former boss aimed at visualizing any type of multi-dimensional data.

The pitch was relatively simple: take any dataset and provide immediate value and visuals to the end user. There are a multitude of ways to tackle this prompt, but my boss had recently come across a hot new data-sci algorithm called t-SNE (T Stochastic Neighbor Embedding). After getting a crash course in t-SNE, k means clustering, and overall dimensionality reduction, we came to the conclusion that a 3D sparse plot would be the way to go.

Running the t-SNE algorithm on a dataset results in X / Y coordinates, so all we had to do was pick a third dimension for the Z axis and we had our 3D cloud. I've always been inspired by the ethereal feel of stars and I wanted the cloud to feel like a glowing constellation. I jumped into learning a bit of 3JS and started hacking a prototype together with good ole jQuery. I had an extra bit of fun working on the intro animation, which I based around the idea of reducing complexity and bringing order out of disarray.

The end result was a prototype that showed really well and generated a lot of excitement. We even pushed forward and started to productize certain aspects of it. The prototype itself was realtime and interactive, the data had just been pre-processed through the algorithm and massaged for the proof of concept.

Star Cloud remains one of my favorite hackathon projects and took a little over a week from start to finish. It was an absolute blast to work on with my boss and I loved getting to learn some new technologies and ways of approaching a problem.