Planting Seeds of Technology

When it comes to start-ups, some people just can't quit. For Judy Estrin, the hardest part about starting a company comes when it isn't just a little start-up anymore--when it merges with another company, is acquired, or goes public and suddenly feels a lot like a big, established company.

Estrin and her husband, Bill Carrico, have been through this a few times: in 1981 with Bridge Communications, which went public in 1985 and merged with 3Com two years later; in 1988 with Network Computing Devices (NCD), which went public four years later; and in 1995 with Precept Software, which was purchased by Cisco Systems three years later.

So what's a person to do?

One could retire. But Estrin, age 46, is bad even at taking a vacation. Or one could take on an advisory role, which she has. Currently she serves on the boards of Federal Express, Disney, and Sun Microsystems (as one of the most prominent woman engineers around, she gets a lot of invitations). But being on a board doesn't make up for not having a start-up to call your own.

Estrin thinks she's finally found an answer: start a company whose product isn't hardware (Bridge made routers and servers, NCD made X-terminals) or software (Precept made software for streaming video), but is instead new companies, grown out of research done internally. That company, Packet Design LLC, Mountain View, Calif., launched last year, released its first start-up, Vernier Networks Inc., also in Mountain View, this spring. (Vernier provides products that allow users to protect, manage, and control wireless networks.)

While the entrepreneurial path Estrin has taken may have initially been unexpected, the fact that she became an electrical engineer was anything but surprising. Her parents, Thelma and Gerald Estrin, are both engineers and IEEE Fellows. The two worked together when Judy was an infant to build Israel's first mainframe computer, the Weizac, for the Weizmann Institute of Science, in Rehovot.

Bred to be an EE

Given their parents' background, it was inevitable that the Estrin children would be exposed to technology early. "I was probably 12," Estrin told IEEE Spectrum, "when my dad first brought home some videotapes on Fortran that he made all of us watch." Of the three sisters, Judy was the most interested. Deborah is now a computer science professor at the University of California, Los Angeles (UCLA), and Margo is a doctor.

"What I liked about math and computer software," she said, "was the problem-solving aspect--which may be what I like about business. It's also about taking this huge problem and figuring out how to break it into pieces and solve the pieces to get you to an end result."

Estrin went to UCLA to study electrical engineering and computer science, then on to Stanford University, California, for a master's in electrical engineering and computer science. There, she was assigned a laboratory project that didn't seem significant at the time but ended up setting her future technical direction. She built a small ring network linking four microprocessors, not a trivial task in those pre-Ethernet days. Ever since, her technical focus has been communications.

At Stanford, Estrin also worked with Internet pioneer Vincent Cerf. At the time he was overseeing a team of seven researchers developing the transmission control protocol/Internet protocol (TCP/IP) that was to be key to today's Internet. (Cerf had done his graduate work at UCLA under Judy's father Gerald.)

Cerf was surprised that Estrin chose not to go on for a Ph.D., like the other members of her family. (Estrin sometimes jokes that this makes her the black sheep of the bunch.) "But," he said, "I have a hazy recollection of her saying that she wanted to go out there and make something happen in industry. Getting her hands dirty appealed to her more than anything academic did."

When Estrin finished her master's degree, she had three interesting job offers: from Hewlett-Packard, Xerox's office systems division, and Zilog.

"I ended up going to Zilog," she said, "because somebody told me the smartest people they knew were at Zilog, and I was better off joining a smart team than a big company."

"I made the decision to go to a small company, not so that I could advance more quickly--though that happened--but because I thought I was going to work with some really smart people. It really molded my career, and my life, since that is where I met Bill [Carrico]."

Early days

In 1976 Zilog Inc., in Campbell, Calif., had about 50 employees. The company had recently come out with the Z-80, an 8-bit microprocessor, and when Estrin joined she was assigned to work on the architecture for the next generation of chips, the 16-bit Z-8000 and the 8-bit Z-8 (which is still used today in embedded systems). Before long, she was made project manager for a local-area-network technology code-named Ariel.

In the mid-1970s, it was unusual for a woman to be given project management responsibility, recalled Eric Benhamou, who worked in the next office at Zilog. (He is now chairman of 3Com Corp. and Palm Inc., both in Santa Clara, Calif.) "But," he said, "she stood out as one of the only people the company could trust to run an important project. She was surrounded by a lot of macho, chauvinistic semiconductor executives. They would literally pat her on the head--she's kind of short--and were always very condescending. But I never saw her lose her temper. I knew she got upset, but she handled it extremely professionally."

Ariel, introduced in 1980 as Z-net, was one of the first commercially shipped local-area networks. Z-net, Estrin recalls, boasted the first operating system that was designed for a network of computers instead of a single computer. The operating system was called RIO.

But Z-net and RIO turned out to be problematic.

The problem wasn't with the technology. "We were pushing the envelope," Estrin told Spectrum. "In those days, memory was expensive, so you were constantly pushing at the edge of memory sizes to build something that was cost effective and had good performance.

"The challenge was marketing. Zilog was a semiconductor company, and here we had something interesting in the systems business. Z-net turned out to be the predecessor to what was to become a huge industry [networking], but the fact that no one has ever heard of it shows that Zilog just didn't know how to sell it."

This changed Estrin's attitude toward marketing. "I used to think that, as an engineer, you didn't talk to those guys," she said. "But this made me see marketing as a field that is important."

While Estrin was project manager for Z-net, a group of Zilog managers went off to start Ungermann-Bass Inc., leaving a management void. Estrin stepped in, and, at age 24, became the company's engineering manager, with some 20 engineers under her. She reported to Carrico, then general manager, who was later to become her business partner and eventually her husband.

"When I joined Zilog in 1979," Carrico recalled, "Judy was recommended to me as a 'smart girl'--not exactly politically correct--and it was suggested that she be my engineering manager." He did move her into that job, because, he said, she was more knowledgeable about computers and networking than he was (having just come in from Fairchild, now Fairchild Semiconductor Corp.), and because she seemed to have a handle on dealing with people.

"Silicon Valley being what it is," he said, "Judy had to manage people who were somewhat difficult." She also had to deal directly with the marketing manager, which wasn't necessarily easy. "There is often tension between engineering and marketing, because the people often do not have the same personalities," he said diplomatically.

Ironically, Carrico said, over time, "some of her biggest contributions [to their various companies] were in marketing. For someone with a computer science background, this ability seemed to come out of nowhere, but really is tied to her people skills, because sales is about having a people orientation."

Zilog, meanwhile, had been fully acquired in the late 1970s by Exxon Office Systems and was starting to feel like a big, politically complicated company. Opting to move on, Estrin left in 1981 to join her former colleagues at Ungermann-Bass. Today, Zilog still exists, as a subsidiary of Texas Pacific Group, which acquired it in 1998.

At Ungermann-Bass, though, she ran into another situation. Charlie Bass and Ralph Ungermann had both known her since she was a graduate student. While they had left Zilog in 1979, she stayed on, gaining two years of management experience. So she was disappointed to learn that Bass and Ungermann still viewed her as an entry-level engineer.

"It wasn't blatant," she recalled, "but I didn't get the respect I needed, and I wasn't able to make the impact that I wanted."

Meanwhile, Carrico, with whom Estrin was now living, left Zilog, intending to start his own company. She quit her job and joined him. The two soon invited Benhamou on board.

Internetworking as a bridge

Carrico and Estrin began talking about ideas for what such a company could do. They settled on the concept of internetworking. A few companies were building local-area networks, but the networks were incompatible. What was needed were internetworking systems--"bridge" systems consisting of hardware and software--to tie different types of local-area networks together. With that concept in mind, they named their company Bridge Communications and, business plan in hand, went calling in 1981 on the venture capital community.

"Raising the money took six months and was very hard," Estrin said. "One VC [venture capitalist] called us up and said, 'Now why do you need $1.8 million to build a modem?' We realized we were not explaining it right, people didn't really understand what local-area networks were about; we had to educate them. The earliness of the market was a big problem."

A second problem was created by her and Carrico's youth and their relationship. "I was 26 at the time, Bill was 31, and people reacted like, 'Who is this guy and his girlfriend?' We didn't have any experience, so what made us think we could run a company?"

At one meeting with venture capitalists at which the point was raised, Carrico finally said, "Look, would it make you feel better if we were married?" The VC said, "No, we know the statistics for divorce rates, we'd rather have the unknown of you two just living together." (Estrin and Carrico finally did marry six years later, in 1987. Her sister Deborah quipped at the time, "They didn't need to get married, they're incorporated.")

In November 1981 the venture commitment came through, but before the money was actually turned over, a technology research firm came out with a report saying that broadband technology, which was then being evangelized by Wang Laboratories (now defunct), was going to win over the 10-Mb/s Ethernet. And Bridge's plan was Ethernet based. "We thought the report was nuts," Estrin said, "but we had to go and convince the investors once again." Meanwhile, Jean Pierre Boespflug had joined the founding team.

Once the company was funded, they quickly hired a team of 10 engineers. The technology development after that was fairly smooth, and the Bridge team demonstrated their product on the day promised in their business plan: one year from the day they got their money.

Bridge went public in 1985 and in 1987 was profitable, selling terminal servers and routers, with 450 employees and revenues of some $70 million annually. Looking forward, however, Estrin and Carrico saw that personal computers were encroaching on the server market, and therefore their products needed PC connectivity. They decided the best way to get that technology would be through a merger, and eventually agreed on a merger with 3Com that retained the name 3Com.

In the management structure, they set up 3Com's Bill Krause as CEO, Carrico as president and chief operating officer, and Estrin as executive vice president and general manager of the Bridge division.

Krause was planning to step down as CEO after the merger went through but changed his mind. That move caused friction when Estrin and Carrico failed to agree with some of Krause's strategic management choices--namely, to focus the company more on building computer systems rather than on developing communications technology.

Although financially beneficial for Bridge investors, in hindsight, Estrin said, merging with 3Com might have been a mistake. "We should have bought Cisco instead," she said with a wry smile. (In 1987 Cisco was a fairly small operation.)

So, in 1988 Estrin and Carrico left 3Com. "They had a choice between fighting it out inside the company, which would have been destabilizing, or moving on. They did the right thing for the company," Benhamou told Spectrum.

The couple intended to take six months off to relax and think about what they might want to do next. Shortly after they announced their planned departure, a group of engineers came to them with a business plan and a prototype for a networked computer, a device later generically called an X-terminal and referred to as a thin-client approach. (An X-terminal acts as a graphical interface to a network; applications run on the server, not on individual terminals.)

"So we took one day off, and then joined them," she said.

"It was a beautiful concept--this was before Windows had taken off, before PCs had dropped in cost."

This time, raising money was a lot easier. The VC firms that had funded Bridge had their checkbooks out in a few hours.

Once again into the pool

Network Computing Devices (NCD), as the new company in Mountain View, Calif., was called, grew quickly into a profitable business. It went public in 1992, shortly before the price of PCs dropped dramatically and Windows rocketed past Unix as an operating system. Estrin took over from Carrico as CEO in 1993 when he moved into a part-time role, and they broadened the company's focus to include X-terminal software and e-mail. She and Carrico both left in 1994.

NCD is now struggling to survive; the drop in PC prices and growing dominance of Windows over Unix-based systems squeezed out the X-terminal business. The company now makes Windows- and Unix-based terminals as well as software for managing thin-client applications over a network.

"I left because I felt that I had the company strategically on path, but I was just burnt out," Estrin said. "We had never taken a break between doing Bridge and NCD. And I had a four-year-old son." Estrin decided to replace herself as CEO and retire. But, she says, "the replacement didn't work out as well as I hoped; NCD's not a fun story to tell after that."

She looks back on the NCD experience as humbling. "When you do an IPO [initial public offering], and look people in the eye, and sell them stock, and two years later the stock is worth much less than you sold it for, that is not fun." But, she said, it may have been positive for her in the long run. "I think you can get really arrogant if you don't have some humbling experiences."

"I made the decision to go to a small company...to work with some really smart people"

At that point Estrin and Carrico told friends that they were never going to start another company and tried retirement.

"We were pathetic," Estrin said. "We spent a couple of months just catching up on life--doctors' appointments, cleaning closets, things I always had excuses not to do. Then we would take long walks just agonizing over what we should do. We thought about getting involved in a number of companies, doing investing. But then we realized that wouldn't work for us. We don't like to suggest things to people and then have them not listen to us. We wanted to build things ourselves."

So from talking about retirement, Estrin and Carrico switched to talking about which technologies were ripe for a new venture. And in March 1995 they incorporated Precept Software, in Menlo Park, Calif., a company into multicast streaming video across a network.

Again, getting funding was not a problem. "For Bridge it took us six months," Estrin said. "NCD took 6 hours. Precept took 6 minutes."

The technology, however, was a problem. From reading the white papers on multimedia applications put out by Cisco Systems Inc., San Jose, Calif., Estrin and Carrico had assumed that multicasting protocols were already being built into network infrastructures. That assumption was wrong. It took several years before multicast actually began to be deployed, eating up the early market advantage they had assumed Precept would have.

"There was a frustration in knowing that you were doing something right, but the infrastructure rollout wasn't happening," Estrin said. And, "by the time the infrastructure started to happen, Microsoft had entered the market."

The only way to compete with Microsoft was to get some muscle behind the product, so in 1998 Precept was sold to Cisco. Precept's software was attractive to Cisco because it drove bandwidth usage and took advantage of added-value features in Cisco products such as multicasting. The sale made Precept's investors happy: the company was started with $7 million and sold for $82 million in Cisco stock, which went up some five times between the sale and its peak in 2000 (and recently plunged with the rest of the Internet market).

Estrin was part of the package that was sold to Cisco, and became Cisco's chief technology officer (CTO), a job she held until 2000. As CTO, she oversaw the mergers and acquisitions teams, the legal and government affairs groups, and a consulting engineering group that dealt with customers. She also led company-wide R&D initiatives, and was charged with driving the company's strategy for investments in technologies.

The corporate big time

This, Estrin said, was the first time she had ever worked for a gorilla. (Precept had 50 employees when it was sold; Cisco had 18 000 when Estrin joined, 36 000 when she left.)

"I'd always worked at small companies where you have to fight to get heard by outsiders, and all of a sudden, there was no problem getting people to listen to you," she said. "The momentum of Cisco was awesome." What's more, Estrin liked being back in the networking business. What she didn't like was the loss of control involved in moving from CEO to CTO.

She also found big company life a little frustrating: "The engineer in me always wants to dig into everything deeply. You can't do that when there is so much going on."

So in April 2000 she announced her plans to leave. She was happy with the job she'd done externally, representing Cisco to customers, and internally, making people aware that Cisco needed to start thinking about software and end-to-end services instead of just selling boxes. "I think I gave a lot in those two years, and I can see my impact in a lot of places, even though there were days when I didn't think anybody was listening to me," she said. "I also gained a lot out of those two years, both in exposure for myself and financially."

"I was surprised she lasted as CTO as long as she did," Benhamou told Spectrum. "She has the intellectual capacity to operate at the highest levels of the largest corporations, but she is also impatient and demanding and likes to have control over what she does. It is hard to do that at a large company; she is happier being involved in smaller companies, where she can control most of the variables."

Again, she recalls, she and Carrico were back to deciding what to do next. "We knew we weren't good at doing nothing. But we didn't want to start another product company."

One reason was that, in the spring of 2000, the venture capital market was what Estrin calls "frothy." "There were all these companies started for the benefit of getting an IPO, not for the love of the technology," she told Spectrum. "And I just don't like that."

She was also tired of starting companies, building teams, and then selling or merging them for various business reasons, not to mention being faced yet again with the struggle of what to do next. Leaving each of her previous start-ups made sense, she said. She just didn't want to do it again.

She and Carrico were concerned, too, that no long- or even medium-term research was being done on Internet technologies. Everyone was in too much of a rush to bring products to market. So, in their best problem-solving mode, the couple put together a business plan that, they hoped, would address all these concerns.

Different rules

Their new company, they decided, would never go public or be sold. It wouldn't be a nonprofit; in fact, it would spin off projects, when appropriate, into separate entities that could be sold. It would hire serious researchers, but pair them with development engineers to keep the organization from turning into a product-less think tank. And it would build close ties with potential customers.

The company, now in Mountain View, Calif., was named Packet Design, in an effort to link it to the Internet (moving packets around), apply its research to a broad range of technologies, and not sound faddish. It was incorporated in May 2000, with an initial investment of $24 million, mostly from individual investors.

This time, Estrin was named CEO from the start. Carrico is chairman. The company now has 29 employees, 22 of whom have technical backgrounds.

As CEO, Estrin sets what she hopes is a fairly sane work example for the company. "I maintain," she said, "that if you're frenetic, you can't do your best thinking. So we're making sure to set a tone that isn't frenetic. Because if you really want people to think, and do things the right way, and not cut corners, you've got to create a different atmosphere."

Estrin typically arrives at her office at 9 a.m., after dropping her son off at school, and is home by 6 p.m. for dinner, though she often works there later in the evening. She occasionally has a breakfast meeting, rarely a dinner meeting. She has lunch with Carrico every day.

The Packet Design team initially identified six projects, with varying time scales, some short term, some slightly longer, some clearly blue sky. All had to do with the Internet infrastructure; a few targeted improving routing within the Internet, one is in security, and one is in user mobility.

This last project became Packet Design's first "product," announced this spring: a company called Vernier Networks Inc. Vernier will sell technology (for which a patent has been applied) to be used with IEEE 802.11 wireless networks. The system authenticates users through a Web-based network interface, and requires no special software for the devices (like personal computers or personal digital assistants) that communicate with the network.

The technology is currently being tested at the University of Washington in Seattle. Vernier Networks has been incorporated separately from Packet Design, and eventually will be taken public or sold. But because the sale will not affect the core company, Estrin in this case will not have to quit just to start something new.