This ex-Yelp engineer just raised $6 million to help make tastier beer

SigOpt, a young startup co-founded and led by Cornell PhD and ex-Yelper Scott Clark, has raised $6.6 million from Andreessen Horowitz to build out its product, which is helping MillerCoors make tastier beer.

Well, sort of. In reality, SigOpt uses artificial intelligence to solve a common, but complicated, mathematical problem - that just happens to be perfect for helping to brew the perfect batch of beer.

You know how it is when you're making a tasty dirty gin martini, and you don't want so much olive juice it covers up the vermouth flavor, but if you add too much vermouth, it totally overpowers everything? It might take a bartender years to perfect their recipe.

SigOpt's whole reason for being is that it takes in all that data, makes an educated guess as to what happens to the whole result if a single variable changes, and makes smarter predictions. In the martini example, you'd tell it you want a strong, but not overpowering, gin flavor, and it'd predict the ideal ingredients and compositions for you to use, minimizing the trial and error.

"We're just trying to get rid of the most tedious part of the job," says Clark.

The same principles have application in everything from stock trading (optimizing the mix of your portfolio for risk versus return) to industrial manufacturing (balancing high quality with low cost materials). Basically, it's software for optimizing otherwise extremely complicated processes.

In fact, Prudential and Johnson & Johnson are SigOpt customers. And MillerCoors is using it to fine-tune their beer development process, reducing the time it takes to find the exact recipe they're looking for.

Plus, SigOpt has experimented with using it for sports betting to beat Las Vegas odds (though you probably shouldn't use it for that).

Optimize everything

The roots to SigOpt go back to Clark's time at Cornell. As part of his thesis, he was trying to do genome resequencing, which turns out to be an insanely tedious process of trying one gene sequence, then another, then another, until something works. It could take hours or days without any kind of result, "churning through government funding" all the while.

Ultimately, it's what Clark calls an "optimization problem." If it doesn't work, you correct the reaction for what you think will fix it, only to find that it has another, unexpected consequence elsewhere. Because it involves complicated biochemical reactions, you can't just trust a machine to get it right, but it's also beyond the ken of most mortal humans to do that kind of math on the fly.

"You can't do 20th-dimensional math in your head," Clark jokes.

Convinced that there had to be a better way than just endlessly and mindlessly entering these sequences, he went around to all his peers. Nope, it turned out, that was it.

So he wrote some software that would help optimize and automate the process of manually trying those sequences, predicting what comes next, and made it part of his thesis.

Post-graduation, and tickled by the idea of staying in computer science, he ended up at Yelp, applying some of those same principles of automation and optimization to how they placed ads on the page, building a popular software tool for doing the same.

He quickly built a name for himself as a scientist in that field, giving talks at eBay, LinkedIn, and Netflix on using math to optimize their products. It was the right place at the right time, as Silicon Valley increasingly turns to machine learning and artificial intelligence, of the kind used by SigOpt, to solve big problems.

After two years of doing those talks, he realized there was no reason he couldn't take his thesis work, combine it with what he learned in Silicon Valley, and build a business applying what he learned to other business problems. He convinced Patrick Hayes, his "smartest friend" at Yelp, to come with him, and SigOpt was born.

Raising the rounds

In 2015, SigOpt entered the famed Y Combinator startup accelerator program, leaving with a little seed money, including a modest sum from Andreessen Horowitz.

This new $6.6 million round is also one of the first investments made by Andreessen Horowitz General Partner Martin Casado, formerly the founder of Nicira, since he left VMware for the firm in early 2016. SV Angel, Blumberg Capital, Data Collective, and Stanford University all participated in this round.

"Given what an exhaustive and costly process it can be, it's little wonder SigOpt is getting the excitement and traction they are," Casado wrote in a blog entry. They get to better business results faster with fewer resources for some of the most salient problems in industry."