Swift, machine learning, iOS/macOS app development, and math.
I currently work here.
The system we've built at HOVER turns 2D images into photo-realistic, size- and location-accurate 3D models. With just a few images we can extract enough information to know the size, type and location of every surface on the building.
A few use cases
Some of my contributions
Glass pulled content from all over the web to show you new releases and deals from your local stores.
As the CTO of glass and the only developer on our 4 person team I wrote all the code.
It wrote a few dozen entries a second to the database while serving up html pages for browsers and json for the constantly up to date iOS application. Glass made heavy use of front end caching, geolocation, asynchronous background tasks and remote API calls.
Glass had thousands of users and was growing steadily. And then we ran out of money. http://www.shopglass.com is offline
Here I built a native framework for both iOS and android that allowed us to get branded applications built and submitted to the app stores in about 72 hours.
The framework included a Rails back end that supported the mobile applications. It was like a CRM/CMS for mobile platforms. With features that took advantage of mobile device capabilities. Some Features include:
Memory Reel started as an application designed to capture your memories. Take a wedding for example. Lots of guests, lots of cameras. Lots of photos ending up in different places. Lots of blog entries and social network posts scattered across the web. Memory Reel collected all of that information and arranged it into a beautiful timeline of your wedding. All the pictures and videos that guests took came together, regardless of where they were posted or hosted.
As time went on we went from being a free service for consumers to being a paid service for professional event planners. As Memory Reel's focus shifted we renamed the company to Moretap and went after the mobile market.
Soon after getting access to my genome on http://23andme.com I started getting reading through the sites forums. I noticed a trend of people asking questions, and then asking what people have for a certain SNP or gene. Things like, what's your orientation and gender? What are your base pairs for these SNPs? Then a bunch of people would answer and they'd try to find a correlation between people's answers to the questions and their genome. A forum seemed like a bad tool for the job so I built DNAqa (dna questions answers) to facilitate such things. Apparently it wasn't as popular as I thought. I shut it down after about a year.
Cloudseye was a crowdsourcing tool that identified objects in mages. The site made it easy and fast to say what you thought was in a picture and the most common answers floated to the top. You could email a picture or send a text from your cellphone and the web application would respond the same way with an answer. It was really neat. You're walking in the woods when you see a strange bug, and your friend wants to know what it is. You have no idea so you take a picture with your phone and send it to cloudseye. A few minutes later you get a message with the top 3 responses from the web.
Soon after launch all kinds of computer vision powered image search engines popped up and cloudseye was doomed. http://www.cloudseye.com