Yunshan Zhu, now a VP at Atrenta, talks about the creation of NextOp, the skill set that made him the CEO and the reasons why they joined forces with Atrenta.
This story begins as many new developments have in the EDA industry. A designer, or verification person in this case, by the name of Yuan Lu, was working in a semiconductor company. He manually developed a lot of assertions for use in formal verification. He found they also were useful in a simulation environment, but he had great trouble convincing his fellow workers to develop similar assertions. He provided training, the engineers attended tutorials, the languages had become standardized, but still the designers remained uninterested and would not even write the simplest of assertions. Yuan reasoned that if he could find a simple way to automatically generate these assertions it would provide a lot of help.
One of Lu’s graduate school friends, Yunshan Zhu, provides this analogy. “One of my kids is attempting to learn Chinese. He loves to read but has problems writing, which is quite a daunting task. You often don’t know where to start. The same is true of designers. They know Verilog and you ask them to learn a new assertion language, where temporal logic is somewhat foreign to them. They have to take the initiative to come up with ideas about what to write. That is a difficult barrier to overcome. When they are provided with an example they can easily see how it works and why it is useful — this is the same difference between reading a book and writing a book.”
Back in the late ’90s, Lu, Zhu and another graduate friend, Armin Biere, were working on model checking algorithms. With the motivation that assertions were useful but too difficult for most designers to write, they decided to try something new and they formed a company called NextOp.
Zhu says that Biere was a really good researcher in both theory and practice. Lu was working in the industry and was very aware of the practical problems. That left Zhu. “It seems as if I was not good at anything so they said to me, ‘Maybe you can be the manager.’” Thus he took on the role of CEO. Zhu jokes, “This is the industry standard, isn’t it?”
“We had some ideas about what we wanted to do, but the manager didn’t have any management experience and the other founders had no entrepreneur experience, so we knew that nobody would fund us. We did get some initial funding from a generous friend who had done well in an Internet company.”
SE: It takes a while to go from a concept to product. Tell us about those early years.
Zhu: This is a fun stage for a startup, and so it is important that you pick founders who get on well together. In 2006 we got a little bit of money — not enough to pay the founders, but enough that we could go to China and find some young programmers. We opened an office in Shanghai, and found some great programmers.
Having an offshore location is more work, and we had to travel to Shanghai quite a lot. If we had not had support from our families we would not have been able to do it. The quality and dedication of the people we found is as good as we could have found in Silicon Valley. When I first came to China to recruit, we had nothing, not even an office. Being from China, I had some credibility and contacts there, and this was important. This became our competitive advantage.
It took about three years to get to product, and in the process we did a lot of iterative refinement. Once we got a customer, we would learn from them and make improvements. Those early customers tend to have a vision match and our first customer had hired people just to write assertions. They knew this was inefficient and were willing to go through the pain of an early product release.
SE: How did you prepare for competition?
Zhu: You need to build a barrier. It is not the idea or the algorithm, because these can be copied. It is the time you spend refining the product and what you learn along the way that is hard to repeat. The use model is very important. There were many things that we thought would be important that weren’t, and so we had to prune a lot from the product and focus our development attention. We were lucky in that we had no competitors, and so we chose our customers carefully and even resisted some of the requests from our teaching customers. For example, one of them wanted a lot of manual guidance, which we tried. We decided that this was not the right way forward because this was not the right way for the majority of designers to adopt us.
SE: What was it like to be the CEO?
Zhu: When you work you often think, ‘Wouldn’t it be nice to be the manager?’ When you become the manager you find that you kind of work for everybody. You no longer have one boss, you have lots of bosses, but what is nice is that you can influence the direction of the company. I have to learn about their needs and work out how to satisfy them.
During the first three years, it was important that I had a technical background because I had to attract the right people to join the company. As we started to engage and work with customers there was a new set of challenges. I had to start working with people within the company that were not as technical, such as sales and finance, and I had to learn how to work with these people.
SE: How did the recession affect you?
Zhu: We had gotten to the stage when we could start seeing customer adoptions and then everything slowed down, but it didn’t impact us that much. As a first-time CEO I had been extremely cautious, so we never really had a high burn rate. As soon as we started to get customers we turned cash flow positive, but the company probably would have developed faster if customer adoption had come earlier.
In 2009 I started to talk to some VCs, after we had some customers. The VCs told us that we needed a professional CEO and that more sales were needed to ensure that they were not just your friends. I started talking to some CEO candidates and continued to get more customers. By the end of the year we had more customers, were cash flow positive, and I didn’t really need the VC money unless I had to hire an expensive CEO.
SE: What was the motivation for talking to Atrenta?
Zhu: Around 2011-2012 most of my energy was around sales. We had two or three salespeople and a similar number of AEs. Our biggest expense was in sales and support. A company such as Atrenta, which is 10 or more times bigger, has a much larger sales force. For us to grow to that kind of size would take a while. There was an opportunity to take advantage of their channel and for Atrenta to have additional products in their channel.
I also believed that the technology would prosper within a larger company and I had to consider getting a fair valuation for the company. Ajoy Bose is a CEO with a technology background, and he understood the concerns that the NextOp founders had. I also had to pay special attention to the AEs and sales folk who would see the biggest change with the change of organization. For the engineering team, the key is motivation. They need to be working on cool, exciting stuff and need to know that the stuff they are producing is useful. In a small company, these things are live or die, but in a bigger company, the R&D team is further back from the users.
It has now been over a year since the merger and the team is still intact. Now I have more time to concentrate on the technology and the R&D team. We are still learning about new use models and adapting the product to fit the channel.
SE: If you had a new idea for a product, would you prefer doing it in a startup or an environment more like Atrenta?
Zhu: There is a big tradeoff. Do you want to trade the efficiency of developing a new product outside of a company? In a startup you live or die by what you do, so engineering motivation is easier. What you lose is the sales channel. This is very expensive to build and takes a lot of effort to recruit. I think it is easier to attract the talent to a startup.
SE: Do you feel, after this adventure, that you have some skills as a manager?
Zhu: That is for my team members to say.
Yunshan Zhu, Ph.D., Vice President of New Technologies at Atrenta.
Prior to NextOp, Zhu was a member of the Advanced Technology Group at Synopsys. He also worked as a visiting scientist and a post-doctorate at Carnegie Mellon University, where he co-invented the bounded model checking algorithm. Zhu did his undergraduate study at the University of Science and Technology of China and received his Ph.D. in Computer Science from the University of North Carolina at Chapel Hill.