Gideon Mendels has spent the past three years in machine learning research, studying at Columbia University, building hate speech detection software at Google, and developing chat analytics app Groupwize with his business partner, Nimrod Lahav. In those roles, Mendels observed a recurring problem. There wasn’t a good system for managing machine learning projects, like GitHub for software or Salesforce for sales.
Mendels and Lahav set out to solve that problem. The result is Comet, a suite of tools that help machine learning teams manage their experiments and projects. Comet tracks code, facilitates team collaboration, and allows developers to glean insights from past work.
“Without this fundamental infrastructure that Comet provides, machine learning teams are inefficient, disorganized, and difficult to manage,” Mendels said.
Comet is based in New York but the team is currently in Seattle for the inaugural Alexa Accelerator, a new program that Amazon is running in partnership with Techstars. The accelerator supports early-stage companies developing technologies related to Amazon’s popular artificial intelligence and machine learning-powered voice platform. Each of the startups in the accellerator pitched at a demo night in Seattle Tuesday.
“Our goal is to facilitate and support machine learning engineering in any company or university,” Mendels said. “We’re building a product that Machine Learning engineers love to use and any organization can’t live without.”
We caught up with Mendels for this Startup Spotlight, a regular GeekWire feature. Continue reading for his answers to our questionnaire.
Explain what you do so our parents can understand it: “Comet allows machine learning teams to automagically track their code, experiments, and datasets creating efficiency, visibility, and reproducibility.”
Inspiration hit us when: “When I was working at Google, I realized that the same pain I encountered doing Machine Learning for Groupwize also exists at the top tech companies. If Google, one of the companies with the best developer practices in the world has not solved this, probably no one has. And indeed, over 100 engineers and managers validated our thesis — machine learning teams are broken.”
VC, Angel or Bootstrap: “We already have one of the top VCs in the world on board and we need investors that can help with scaling our company to meet the demand. The global spend on AI (which is powered by machine learning) is already at $40B and according to Pwc, would increase the North American GDP by $3.7 trillion. It’s hard to bootstrap a company to seize such a huge opportunity.”
Our ‘secret sauce’ is: “A solution that works with any cloud provider, any Machine Learning software, and for every task (text, speech, vision, robotics etc).”
The smartest move we’ve made so far: “Joining the Techstars Alexa accelerator. Our managing director, mentors, and advisors have played a crucial part in our success.”
The biggest mistake we’ve made so far: “Not working from the same office earlier. When we started, I was in New York and Nimrod in Tel-Aviv. Although we did manage to make a lot of progress, we realized that working from the same office is much more productive.”
Would you rather have Gates, Zuckerberg or Bezos in your corner: “Bill Gates – We both grew up using Microsoft tech and have very fond memories of hacking early windows versions. On a more serious note, Gates is a mix of a highly talented engineer and businessman.”
Our favorite team-building activity is: “We love to hit the gym and play video games together (Currently into Fortnite Battle Royale).”
The biggest thing we look for when hiring is: “Cultural fit: At this stage of the company it’s very important to hire people that are comfortable with our culture. We love to solve problems quickly and without reinventing the wheel.”
What’s the one piece of advice you’d give to other entrepreneurs just starting out: “You’ll get a lot of advice from people — while some will be supportive, you’ll also get a lot of negative feedback. While it’s important to absorb that feedback and really think about it, eventually you have to listen to your inner voice guiding you to the right path. You know your company better than anyone else.”
Editor’s note: GeekWire is featuring each of the nine companies in the Alexa Accelerator leading up to their Demo Day Oct. 17.