Machined Bells
2018Neural networks (self-organizing maps) conduct a 64-voice granular synthesis engine, exploring the sonic and symbolic world of bells.
I’m a composer and violinist who builds interactive systems, and a data scientist who puts machine learning to work for mission-driven nonprofits. This site is where those threads meet.
Selected work
All work →Listen
All music →Recent writings
All writings →Product Onboarding New Users: Measuring the Promise
What does a new user think that your product, or any product, promises to do for them? On a recent multi-state drive I had the opportunity to listen to hours of radio advertisements (thanks to my 6th grader’s insistence on top-40 stations only) and got to hear many “promises.” “We’ll give you cash back on gas purchases,” “We will protect your privacy when you go online,” and “Find the lowest prices on a new car.” However, what the marketing team thinks is the promise is often not what I, the consumer, understand it to be. In buying our mini-van we didn’t need features or lowest prices, we needed reliability, confidence in the safety of the vehicle (precious cargo onboard!), and enough inner capacity to easily pack everything we could need for a weekend trip to the relatives or camping. We ended up paying more for the vehicle that satisfied, and delivered on, the promises that we were looking for.
Measuring the Right Things, pt. 1: What do people click on?
People do not use our products and software in the way that we, the designers and builders, initially expect. In analyzing how web pages are used I wanted a quick way to determine what people are actually clicking on. We have web analytic tools that tell us total page views, link clicks, and form submissions, but is there more we should be looking for? And I had a hunch that users might be clicking on non-interactive elements, expecting something to happen, and experiencing more confusion than joy.
Best Practices in Data Pipelines
The methodology of pulling data out of one software system (via some sort of application programming interface or API) and putting it into a database or other system is commonly referred to as an Extract-Transform-Load (ETL) process. There is also advocacy for ELT patterns (transforming last), but regardless of the location of the T I’ve found that the E and L parts are the most critical. In terms of operational efficiency and the need for constant maintenance, the acquisition of data (E) and storage (L) have to be bug free and running cleanly.
Currently residing in Stamford, NY. Frequently found hiking in the woods with his family, or flying kites on windy days.