Founding a successful AI and machine learning company is no walk in the park. It takes graft, commitment, creativity, and improvisation. This April, as we celebrate the anniversary of our seed funding, it’s the perfect time to give our Opsani users and followers the fun, untold story of Opsani’s early years. 

Ross Schibler and Peter Nickolov began talking about starting a business together. Both engineers are serial entrepreneurs with a number of successful exits under their belts. At the time, the use cases for autonomous technology were just coming into focus, and the two felt that’s where the next big opportunity was going to be. 

Finding Product/Market Fit

What wasn’t as clear was where they’d apply AI and machine learning, the foundation of autonomous operations. As an entry point, they tried to train machines to run themselves more efficiently, and to do that they needed lots of data. Ross and Peter harnessed 200,000 servers and started collecting information. 

After some experimentation and thorough analysis, the pair came to conclude that automation through AI looked to be a good solution for streamlining the CI/CD DevOps processes. 

The Serendipity of Proximity

Up until now, the company had been distributed, with Ross in the Bay Area and Peter in Los Angeles. To make things more efficient, Peter moved his family to San Francisco. The property included a pool house which became the fledgling company’s office for the next six months. 

Ross began studying neural networks, the technology autonomous cars use to survey their surroundings and safely keep operating. During one lunch, Peter complained to Ross about a former employer’s messy way of setting up applications after they came out of deployment, which sparked an idea: could the neural networks Ross was studying autonomously tune software performance in their runtime environments as they do with autonomous cars? 

Shock and Awe

Using neural networks to find the best combinations for application performance and cost-efficiency could be the way. Ross and Peter decided to use autonomous technology to continuously optimize runtime environment applications. By the next month, Peter had built the prototype. 

To test the new code, they first tried it on their initial product, the big data application, which had been running for more than two years and continuously tuned by a number of smart engineers. If any app was already running efficiently, this one was. Peter and Ross anticipated they might see a 20 percent increase in performance with the new software, which would be pretty good but not really a business. If the AI could get more than 40 percent better results, then they knew they were on to something. 

The AI and machine learning tool took a few hours to ‘learn’ its surroundings and then went to work finding better and better runtime combinations for the application and its surrounding environment. Peter and Ross set the application up and left for the night. In the morning, they returned to find the new AI application had tuned their five container server app more than 170 percent, which was stunning, and probably meant the new program was broken. But after studying the results, they confirmed that the AI was not broken. The optimization solutions that were so efficient and effective that not even humans would have thought to try them. 

Proof of Concept

Ross and Peter’s technological success attracted the attention of seed investor Zetta Venture Partners’ Jocelyn Goldfein, an AI and machine learning technology specialist. Knowing total cloud spend for 2020 was going to be north of $50 billion and growing 20 percent year over year, and the potential for companies to save millions of that spend by using this AI to tune their applications was enough for Jocelyn to fund the building of Opsani’s revenue product. 

After a few deployments of the revenue product, including massive scale operations like Ancestry.com, Opsani received a $10 million Series A investment to expand the business and become the solution to a fundamental problem in DevOps. Most organizations put more emphasis on the code rather than the run-time environment in terms of parameters such as VM instance type, CPU shares, thread count, garbage collection and memory pool sizes. This restricts their ability to run applications in a faster and more cost-efficient manner. Using neural networks, AI and machine learning to continuously optimize the application after delivery, Opsani can deliver a better than 40 percent savings on cloud costs, and more than 200 percent better application performance for the end user.

Over the space of two years, what had started as an idea to bring autonomous technology to DevOps wound through a number of twists and turns, starts and stops, and serendipitous events to become a new best practice for the entire DevOps toolchain: Continuous Optimization.

On the anniversary of our funding from Zetta Venture Partners’ Jocelyn Goldfein, we would like to say: Thanks to everyone who has supported and followed us along the way. We’re just getting started!