Two decades ago, a young job applicant walked into IBM’s research center in Brasília for an interview. Told that Big Blue couldn’t afford to have him running off to do a mid-career PhD, he left convinced that he’d have to pass up perhaps the most desirable IT research gig in Brazil.
Now 43, Henrique Malvar is too old to be called prematurely bald. But he’s spry and fit enough to seem a bit precocious as a top-flight manager in a heavyweight corporation. Malvar serves as senior researcher and manager of Microsoft’s Signal Processing Group. His unit explores ways to compact and transmit data. That puts Malvar on the cutting edge of the quest for a technological fix that should pull the Internet out of the rut of its text-dominated present and catapult it into a more user-friendly multimedia future.
Wearing dark slacks and sweater, sans necktie, Malvar skips across the courtyard from one nondescript office park building to another in quest of a discounted latte. Cut-rate coffee at the in-house Starbucks counts among the perquisites for Microsoft employees in suburban Redmond, just a short jaunt over a floating bridge from Seattle. Having adopted the dress code and requisite caffeine habit, Malvar would seem a true denizen of the New Economy army in the moderately hip and understated Pacific Northwest.
Not for this transplanted carioca the melancholy saudades said to so often afflict Brazilian expatriates. He’s too engrossed in his work. Even in the midst of an economic downturn, Malvar frets little about having to scale back his own work or that of his group. “Bill Gates himself thinks that research is important,” he said. “He’s not thinking about making cuts.”
Smells like nerd heaven.
This is the story of how Malvar got there. It is the story of a self-confessed geek – the father of an arcane but scientifically significant concept called the Malvar Wave. It is the story of a man who is not afraid to take risks, of someone who applied some of that celebrated Latin American resourcefulness to make his own breaks.
In quest of that coveted PhD, Malvar dragged his young wife from her Brasília home to the Boston area, to MIT, one of the world’s top universities. “I really wanted to get my doctorate,” he explained in his office a few blocks away from the main Microsoft campus, with a sprawling layout that discourages pedestrians in the spirit of the University of São Paulo (USP) or Brasília itself. “I always saw myself as a scientist.”
Seven years later, degree in hand, Malvar returned to Brazil to accept a position on the faculty at the University of Brasília. He put together a digital signal processing research group and turned it into a little dissertation factory. He and some colleagues mounted a prototype soundboard, a clunky thing that today sits like a prehistoric relic on a shelf in Malvar’s office. As Brazilian academic careers go, Malvar seemed to be on the fast-track.
But fast-track to where? Dubbed Rico by family members simply because they needed to distinguish him from Henrique the grandfather and Henrique the father, Malvar’s professorial salary hardly allowed him to live up to his nickname. Two attempts at entrepreneurial moonlighting failed when his start-up companies ran aground in bureaucratic shipwrecks. With two small children, he was struggling with a dilemma faced by many Brazilian parents: academically inferior public school or financial debilitating private education. Like many Brazilians in the post-dictatorship era, he was coming down with hyperinflation fatigue syndrome.
About this time Malvar received a call from a former MIT classmate. Though bright, Brian Hinman always seemed too preoccupied with the entrepreneurial opportunities to focus on his studies. He dropped out of the PhD program. “There’s money to be made,” he’d tell his more studious Brazilian colleague. “When you get your doctorate, come work for me.”
No kidding. Malvar wasn’t exactly surprised when Hinman’s voice came over the line to launch into a recruiting speech for PictureTel, a video conferencing company. PictureTel needed to compress its television signals to cut transmission costs and ensure profitability. They needed a signal processing guy. Malvar accepted a position as vice-president of research.
Profitable PictureTel became, but before long the company faced a classic challenge for many high tech innovators. Its once premium service began to devolve into a mere commodity. Margins fell. Malvar found himself laboring over business plans and fighting over budgets. “Where’s the real stuff?” he asked himself.
Tooling around at home one weekend, Malvar thought about Microsoft. This was four years ago, and the software giant’s research center had begun to rival the legendary Bell Labs and IBM Research, the world’s premiere private sector IT centers. Yet the firm had no Signal Processing Group. Malvar sat down to compose an e-mail message to then-Microsoft Chief Technology Officer Nathan Myhrvold. "You don’t have this unit and you should,” said the message. He clicked send on Sunday afternoon.
“By Monday morning they’d made a decision,” he recalled, still astonished at the power of a single, well-aimed e-mail message. In short order, Malvar was running the spanking new Signal Processing Group at Microsoft.
These days about half of his time is dedicated to management duties, but that counts advising and guiding an international group of junior researchers. Malvar only chalks up 5-10% of his time to purely bureaucratic chores. “The image is of Microsoft as empire and Bill Gates as emperor,” he said. But only three layers separate Malvar from Gates on the organizational chart. “Microsoft is like a big incubator under an umbrella,” he countered.
Like most geographically wayward Brazilians, Malvar sometimes dreams of returning to the fold. But not to start a high tech company – at least not in today’s regulatory climate. “I’ve already founded two companies. Man – what a hassle,” he recalled. “The government puts up roadblocks when it should be helping you.”
So for now Malvar’s just pleased that his daughter plans to enroll at the nearby University of Washington instead of going away for college. Besides, he’s having too much fun riding those Malvar waves in Redmond.
The Legos of Success
Stacked in the windowsill of Henrique Malvar’s office are two dozen cubes. Except for their inscriptions, they might be mistaken for children’s building blocks.
The inscriptions commemorate patents for which Microsoft has applied based on Malvar’s research. The company hands out such mementos whenever one of its employees makes a patentable discovery. If Malvar’s pile seems kind of big after just four years on the job - well, it is. “Our batting average here is a little better than elsewhere,” said Malvar, slipping easily into gringo baseball lingo to describe the success rate of his Signal Processing Group.
Among Malvar’s base hits, for example, was the original architecture of something called the MSAudio codec. “Codec” stands for compression-decompression. That in turn is shorthand for a two step process by which digital data is shrunk for quicker and easier transmission and then rebuilt or enhanced on the receiving end to ensure original quality for the user.
In preparing their data, or signals, for transmission, Malvar and his team had to pare extraneous fat like an old-time neighborhood butcher carefully trimming choice cuts of meat. A digital audio signal contains sound data that extends beyond the human ear’s capabilities. Since nobody will tell the difference anyway, all that stuff can be snipped out.
Malvar handed his MSAudio codec over to Microsoft’s Windows Media product teams. They transformed it into something called Windows Media Audio (WMA). WMA files are about half the size of MP3 files with the same level of quality, Malvar insisted.
Malvar’s signal processing group is also working on technology that would allow for web telephony sans headphones and microphones by filtering out feedback and background noise. Another project would make non-text data, notably audio and video, searchable on engines like Google. Yet another may finally solve the problem of video transmission over the Internet.
Since he wins little cubes with each success, perhaps it isn’t surprising that Malvar uses a building block metaphor to describe his work. “It is like building something with legos,” he said. He and his researchers scour the academic literature for pieces of the solution. Usually they find most of what they need, but it isn’t always obvious how to fit one piece to the other to make a neat little building. And then, usually, a key piece is missing. “You have to invent that piece,” he said. “And that could take weeks, months – or years.”