My name’s Michael Baldwin, and I’m a web-focused product manager, full-stack developer, architect and overall entrepreneur.

I have a deep technical background in programming and creating websites, but also a strong eye for design, and the product and people skills to lead cross-functional teams and to realize clients’ visions.

In addition to the long list of websites I have created and managed for employers, I have also independently built and launched five major website projects, leading to front-page online coverage in Wired and The Wall Street Journal, as well as coverage in The Economist online.

I live in New York City, and am currently the Senior Product Manager of Adaptivity at Knewton. Prior to that, I founded Flip Languages and worked with the founders of OkCupid at OkCupid Labs. You can view my portfolio of personal projects below.

show more images (6) » hide extra images «
bootstrapped startup 2012–present
Stats Over 100,000 source lines of original code More than 120,000 words of original learning content

After having taught English as a foreign language in Brazil for several years, and having become fluent in Portuguese as well, I found myself in the unique position of having both a deep understanding of language learning from multiple perspectives, as well as a deep background in programming.

Frustrated with the inefficiency and inadequacy of existing language-learning solutions (both classroom- and computer-based), I developed a new method for computer-based language learning based on spaced repetition, statistical corpus research, language dependency modeling, integrated grammar and vocabulary, high-context practice content, and in-depth explanation. I designed and programmed the entire system, developed its content, and turned it into a commercial product.

Flip teaches English vocabulary to Brazilian students at basic, intermediate and advanced levels. Flip was launched in private beta in mid-2013, and began generating revenue upon its public launch in late 2013. It is available as both a desktop and mobile web app, and operates on a freemium model.

show more images (6) » hide extra images «
independent project 2010–present
Stats Over 3,000,000 completed surveys More than 200,000 Facebook likes More than 1,000 visits each from over 90 countries

Accurately measure the size of your English vocabulary with this quick five-minute test. Originally a weekend project I built “for fun,” traffic on it exploded worldwide, and it is now the most popular tool in the world for English speakers (both native and foreign) to measure their vocabulary size.

In an associated research blog, I show and discuss resulting correlations between vocabulary size and age, SAT score, and much more.

Site growth has been entirely viral, without any major media coverage. It generated one of Hacker News most popular submissions of 2011, with over 300 upvotes and over 300 comments. It has also been featured twice in one of The Economist’s blogs.

show more images (9) » hide extra images «

Crap, I missed it!

no longer active
independent project 2008–2009
Stats Had over 10,000 active subscribers Nearly 1,000,000 e-mails sent Over 95% subscriber retention rate

Crap, I missed it! was born out of the frustration I felt when I missed a favorite artist’s concert, just because I didn’t know it was happening. I knew I could build a service that sent out e-mails about concerts, but then realized the concept could be expanded to a whole host of other areas: new CD’s from the same artists, new books from favorite authors, new TV episode dates and times, upcoming sports matches, currency exchange rate swings, and even lunar and solar eclipses.

I built an extensible notifications platform that could take inputs from a variety of sources (various web API formats, screen scrapers, and manually edited data), store event data in a unified database format, and send out customized reminders to users in a single daily consolidated e-mail. I then designed and implemented the website itself, gave it the name Crap, I missed it!, and launched. An early article on Lifehacker generated thousands of initial signups.

Technically, the site's most groundbreaking feature was a novel method for uploading a user's 100+ MB iTunes library file (done in order to extract artist names for bulk concert and CD notifications). To limit resource use and provide immediate user feedback, I custom-built a lightweight independent web server, which immediately extracted all relevant data as the file was uploaded, without utilizing disk space. The server then instantaneously returned extracted data to the user, who could immediately see artist names popping up in their browser during the upload itself. This cutting-edge technology was featured in an article on Ajaxian.

The site was one of the very first e-mail event notification services, and turned out to be amazingly “sticky” — less than 5% of the 10,000+ users ever unsubscribed. Unfortunately, in the end it did not turn out to be economical to maintain. With a monetization strategy based mainly on Amazon affiliate links, it would have required a vastly larger number of subscribers to justify the time and effort necessary for maintaining the various data feeds and adding new ones, and a viable growth strategy was not found.

show more images (9) » hide extra images «


no longer active
angel-funded website 2007–2008
Stats Over 70 dynamic page templates Scalable across up to 64 MySQL database servers Over 700 KB of pure PHP, HTML and CSS code

ZotFish has been one of my two most ambitious projects so far. It allowed users to submit questions to specific public figures (politicians, artists, celebrities, etc.), other users to vote on those questions, and public figures to be able to respond to those questions.

The inspiration came from a desire to harness the power of a site like Digg, where tens of thousands of users vote on content, but to use this to allow the public to speak with public figures on an equal footing — to give the public a “voice” with which it could question and/or hold accountable politicians, writers, artists and more.

The site was an entire social network platform built from scratch, which allowed users to create profiles with pictures and biographies, submit questions to public figures, vote and comment on other questions, and even vote on other people’s comments. Then, verified public figures could respond directly on the site, either via text or video, and users could further comment and vote on the response. A full admin interface was provided as well.

Upon its launch, a very positive full-length review on Mashable concluded that “not only does [ZotFish] sound fast and easy, but as long as there’s participation, like Digg, there’s no reason why it shouldn’t work.”

However, participation turned out to be the classic chicken-and-egg problem — I couldn’t get more than a couple hundred questions without public figures already committed to participating, and I couldn’t get public figures to participate without already demonstrating a much wider user base. I knew this would be the central gamble of the site, and in the end it didn’t pay off.

But in the process, I gained invaluable experience in large-scale web development, built upon principles like horizontal database partitioning (sharding), memcache usage, Amazon S3 for image hosting, and much more. The database and front-end code was designed from the ground up to be able to scale up to hundreds to web servers.

While no current site duplicates ZotFish's functionality, many of its ideas have become widespread: Twitter provides a place for public figures to communicate directly with the public, and sites/products like Reddit's AMA (Ask Me Anything) threads, Digg Dialogg, and Google Moderator allow users to vote up popular questions in one-off events with public figures. While ZotFish itself never took off, the ideas behind it certainly have.

show more images (6) » hide extra images «
independent project 2005
Results Over 60,000 survey participations Lead story on Wired.com Front-page feature article in The Wall Street Journal online Included in textbook Human Geography Shown in documentary How the States Got their Shapes

The CommonCensus Map Project is an Internet survey I created which asks people about the various geographical affiliations they feel about where they live. It then uses these answers to draw maps of self-identifying regions. Essentially, it visually answers the question: if we had to start over from zero, and draw new lines for states, cities, and neighborhoods, what would they look like if it were done in a democratically “fair” way?

Since existing mapping software could not accommodate the algorithm I designed, I wrote a program from scratch which analyzed the tens of thousands of survey responses and distilled them into meaningful maps, calculating average regional responses per-pixel and discovering the boundary lines between neighboring areas.

To boost the popularity of the site, I also added a sports component, which shows in detail exactly which areas of the country have what proportions of fans for national football, basketball, baseball and hockey teams.

By now, the site's graphic design is quite dated, and a truly accurate map would require millions of survey responses, but the site continues to serve as a valuable proof-of-concept of the algorithm. And while it is no longer being actively maintained, results from the site have now become a reference in understanding human geography. It is used by college professors in geography classes across the country, is featured in the textbook Human Geography, was recently shown in the documentary How the States Got their Shapes, and was the subject on front-page articles on Wired.com and The Wall Street Journal Online.