Background
With my project partner, Graham Jenkin, we explore how live data sources, predictive models, and large language models (LLM) can be integrated to produce compelling content for many domains. We have developed expertise in prompt authoring and management, AI data operations, and delivering a fun and quirky consumer experience. Problem It's less a "problem" than a sports entertainment challenge: how can I make sure I don't miss out on the most interesting games and events for teams and sports I care about? And, can AI be used responsibly and creatively to create an editorial point-of-view that is worth its subscribers' time and attention? Approach Being deeply smitten sports fans ourselves, we decided to start Tuze Sports as a collection of game previews and result predictions for the world's most popular pro and collegiate sports. Albert, our first AI-generated site personality, writes matchup previews for the web with its own irreverent house style. The system also sends weekly/weekend email digests covering "hot" matchups as well as specific teams that fans follow in their accounts. We collect direct customer feedback on the site, and we prompt our community with occasional surveys to solicit new content ideas and other suggestions. We have built Tuze with a mobile-first design approach, so the site is consumable on all modern browsers and device form factors. Outcome This is an active project, so the outcome is ongoing. We consider Tuze to be a lab experiment in audience engagement with entirely synthetic (AI) content and potential monetization. Can we create content relevant and fun enough to merit sponsored advertising and/or subscription revenue? Other experiments: Choose your own tone: let subscribers tune the "house style" to match the personality they want delivering their updates. Can personality be a 'moat' to promote exceptional content quality? Expanded editorial formats: write game summaries and recaps, and even summarize other editorial and social media activity taking place around followed sports and teams.
for upcoming matchups in almost two dozen sports.
Background
I joined Spare5 as Chief Product Officer just as it spun out of Madrona Venture Labs in 2014. Spare5 applied crowdsourced human insight to enterprise data organization, labeling, and rating tasks. The company eventually rebranded as Mighty AI, keeping Spare5 as the community app brand. Problem Companies have lots of data, but it can go stale, become disorganized, or otherwise lose relevance and value. Or, in the case of companies with computer vision features (example: self-driving car camera systems), they need a large volume of labeled data to train neural networks. Labeled images train AI for autonomous vehicle computer vision systems to ”see” the world the way we do and to enable autonomous driving. We also attracted a large number of Central and South American users, so we needed to offer Spanish language content as a first-class experience rapidly. Approach We built a two-sided marketplace: people from all walks of life downloaded our iOS app (Spare5), and we paid them to supply their skills and insights on one side. On the other side were companies with data sets and the need for all of that human skill & insight to improve their data. We turned the data into surveys, games, and rate/review tasks you could complete on your phone. Over time, the company focus landed on enabling Spare5 users to label vehicle camera sensor images. We realized that complex data labeling tasks would only be scalable and feasible on the web, so we had to design a new web-based labeling workflow, while maintaining a consistent UX around account management, weekly payouts, support requests, and task training/educational materials. Under my direction, the product design and community operations teams undertook the following practices: Developed all desktop and mobile experiences with localized language support and accommodation Designed game experience features (levels, experience points, and a prototype for live team-style collaboration) to help quantify labeler performance and provide incentives to practice and improve skills Conducted regular usability tests of new features pre-development Hosted bilingual webinars to introduce new features, demonstrate more complex labeling tool best practices, and collect ongoing feedback Outcome The Spare5 app was never rated below 4.4 stars on the App Store for the ~4 years it was live We successfully transitioned all labeling activity from mobile to web over a ~16 month period, leaving our mobile app with a single paid photo-taking task suitable for smartphones
(I shot the live clips and post-produced the rest using Camtasia.)
BONUS: Delicious Mobile Prototype!
In 2024 I've consulted with MOTO Pizza, an up-and-coming Seattle pizza purveyor, about their digital strategy. MOTO makes wonderful "odd pizza"—check it out, you'll be hungry the moment the page loads— and one of the places you can get MOTO is T-Mobile Park, home of the Seattle Mariners. Problem But the lines to order can get long—like, really long—and waiting for pizza instead of watching baseball and eating pizza is nobody's first choice. Approach I proposed a mobile, game-day-only order ahead feature, powered by a custom app. The app could expand to other delivery and order use cases outside of the ballpark, of course, but start here at this known customer pain point. Try out the prototype iOS experience »Moto Mobile Outcome TBD! The prototype is as far as we've gotten so far. But the prototype has been well-received.
Background
I worked for AngelList in 2014, which is a fundraising and talent hiring platform for startups. One investor in AngelList, GV, also happened to house the design team of Jake, John, and Braden, who are better known now as the authors of the book and process "Sprint: How to Solve Big Problems and Test New Ideas in Just Five Days". Problem I was developing a concept for a mobile app to help better connect qualified job seekers with startups wanting to hire tech talent, all on AngelList's platform. But there weren't any engineering resources available to start building a new app unless the business case could be made much more concrete. Does an app like this help startups seeking to hire talent to screen, interview, and hire the best candidates faster? Approach The GV team offered to help and we decided to use Sprint validate the concept over just five days. You can see the "Sketch", "Decide", "Prototype", and "Test" activities under way in my photo gallery below. Source: The Design Sprint In "Decide", we concluded that the prototype should feature the job seeker experience. We concluded that it would be most valuable to see if we could attract great candidates with an experience that rapidly matched them with exciting startups that aligned with their goals. Increase hires, increase referrals to the platform as the place to go for tech startup jobs. In addition to participating in all of the Sprint phases, I conducted the Friday "Test" activity. I managed all of participant recruiting, conducted and video-recorded each individual test session, and debriefed with the entire Sprint team after all the feedback was in. We selected test participants by searching through an existing list of job seekers on the platform who were in the San Francisco area and available to visit our downtown SF testing location in person. We let five participants experience the "1XUX" - first time user experience - of discovering, onboarding, and making their first match with an employer. I conducted the ~1 hour sessions, giving job seekers task prompts and reminding them to think aloud so the rest of the team viewing the video live stream of the test could take notes. Outcome We concluded the concept had merit, but that more details about the startups job seekers located and getting the nuances right about text/phone/video intros between candidates and startups would be critical. Setting realistic expectations for seekers and startups about how they would connect appeared to be the least developed prototype detail of most actual importance. In the end, I did not get the CEO's buy-in on the app concept, despite it testing fairly well with potential job seekers. Lesson learned (and applied!) in multiple sprints I facilitated later: make sure the key product decisionmaker is in the room early, and often. Had the CEO seen the original pitch and then the actual tested prototype they would have been likelier to be invested in giving it a real trial run vs. just seeing the summarized test findings. This first time through Sprint taught me how to facilitate (this was several years before the book was in print and went to #1 on the NYT Business Bestseller list) - from the O.G. inventors of the process. It was a delight. It's still one of the most powerful design methods I can apply to any early stage uncertainty about a possible product direction.
Background
FeedBurner (my third startup in ten years with the same founding team) helped bloggers, podcasters, and big publishers measure and manage their content's distribution in the pre-social media era. Podcasts still rely on the syndicated feed standard to update pod players when new episodes drop. I was there as Chief Design Officer from the time my co-founder Eric pitched the FeedBurner concept over lunch through to the day Google closed its $100M acquisition of the company. The greatest single skill I honed during this four-year-run was to trust, delegate, and hold accountable other design team members instead of assuming I had to do it all myself. Product + UX I led a three-person design team responsible for most site copy, visual design, front end development, all workflows, and information architecture. I also led several usability evaluation research sessions, which involved bringing local Chicago-area FeedBurner users to our offices to test new features and qualitatively identify design flaws in upcoming designs. I designed and coded the front end of a "bulk management" version of FeedBurner for publishers with dozens/hundreds of publications, called FeedFoundry. I also designed the "flame" logo and text lockup. UI / Front End Development I wrote much of the HTML/CSS/JS in close collaboration with the other two designers, and dabbled in middleware (Java Struts/Tiles). We primarily used the then-brand-new JQuery framework to simplify front end feature development. I learned the hard way why you never deploy code on Friday night. 😩 Service Design I oversaw design and daily operation of publisher community support resources. FeedBurner had a loyal publisher following, so we were able to host lively public message boards and regularly engage with superstar volunteers who helped us support a large community. We also offered direct email support and built a culture of "everyone in product engineering takes a support rotation". We were well known for responsive support.
If you just can't get enough of FeedBurner, well, by golly, here is some more.