PDX Tech Talks Monthly #7 - FutureTalks at New Relic

Dearest reader, I have neglected my duties to write about PDX Tech Talks for many months. Have I been attending tech talks? Yes. I’ve made it to a few of my favorites like PADNUG and AgilePDX, but didn't ever get around to posting the summaries. Why so busy? Well, in the fall I was heads-down hacking on an IoT side project. Since then I’ve been hard at work helping to build some teams at Surescripts. Now that we’ve filled most of our newly-built office space and assembled some great teams, I can get back to some leisurely blogging.

The FutureTalk series, sponsored by New Relic, has been on my radar for a while. When I noticed this month’s talk was about Next Glass, a start-up that uses software and science to make great beer recommendations, I decided it was time to go check it out.

The FutureTalk crowd at New Relic

The FutureTalk crowd at New Relic

The Event – FutureTalks at New Relic, with Next Glass

New Relic, a Portland and San Francisco based start-up that makes software analytics products, hosts their FutureTalks series monthly in their downtown office. The talks are given by tech luminaries like Ward Cunningham, Amber Case, and DevOps guru Gene Kim. The presentations focus on “examining the code of tomorrow, celebrating and meeting the people who write it, and looking ahead to the future of modern software.” Their shiny new workspace is up on the 29th floor of the Big Pink building downtown, and features amazing 360-degree views of our fine city. I could see my neighborhood from up there! I wandered around a bit afterwards and discovered a very long table full of Legos to play with. Awesome :)

This month’s presenter was Trace Smith, the COO of Next Glass, who made the trip all the way from Wilmington, North Carolina where their start-up is headquartered. Trace recounted the founding story of their company, which strives to make wine and beer recommendations that are accurate and personalized. They do this by rigorously analyzing wines and beers (over 35,000 so far) then creating a “DNA profile” for each product.

Who I Met

It was a full house at New Relic, with plenty of folks to chat with. I met quite a few students and alumni from Epicodus, a 4-month program that teaches coding skills. One recent Epicodus graduate, Natalie Blackburn, mentioned that she really enjoyed the program and is now busy applying to entry-level dev positions. I also met Paul Burkett who had made the trip down from the ‘Couv, where he is CEO of the custom software shop AgileCore. Paul was full of interesting tech ideas, some involving Google Glass, and was excited about future entrepreneurial possibilities.

What I Learned

I’ve been reading a lot about the role of “data scientist,” a job title that scarcely existed three years ago. It’s gaining popularity as online retailers and service providers (e.g. Amazon, Netflix, Pandora) strive to make use of their hordes of data. They want to recommend a product, song or movie to you that you’re statistically likely to love.

Next Glass Chief Science Officer Connell Cunningham 

Next Glass Chief Science Officer Connell Cunningham 

Next Glass is definitely riding this technological wave. They added a CSO (Chief Science Office) to their team in addition to the traditional CTO role. Like Pandora's Music Genome Project, the first thing Next Glass did was devise a way to create the “DNA profile” for beverages. Now, I’m no biologist, but it sounded like they used something called a mass spectrometer to analyze the wine. Apparently it uses Thermo Scientific Orbitrap technology and records 22,000 data points for each wine sample. Here’s the “for dummies” video about how it works. After building their “Genome Cellar” database, they started testing recommendation algorithms, comparing them with user taste tests. Users get to rate beverages on a four-star scale, and their preferences are thrown in the data analysis too. The end result: a wine suggestion that users agree with 86% of the time. Not too shabby!

The demo of Next Glasses app was very cool. Interestingly, they ended up using algorithms that work super-fast on a mobile device, despite being about 2% less accurate than the ideal but slower-running versions. Speed matters to mobile users. For me, the most impressive part was their augmented reality feature. While using their app, you could point at beer label with your phone’s camera, and their recommendation rating would hover over the image on the screen, allowing you to click and drill down to find more info. They showed their app scanning a restaurant wine list, then transforming it into a formatted page with hyper-links, high-lighting the wines and their recommendation rating. “Oooohhh,” I said. Very impressive indeed.

Wrap Up

I definitely recommend checking out the next FutureTalk at New Relic. By my estimate you are 87% likely to enjoy it, and I don’t even need big data analytics to back that up. You also might like the Next Glass app, which is scheduled to launch around October.

Next Month

Keeping on the big data topic, I’d like to check out the Portland Big Data User Group. I’m also interested in learning about Docker, and it looks like there’s a local meet-up for that too. Neither of those appear to have anything scheduled yet for September though. If you have any recommendations for next month, please let me know!