I'm at the ACM-hosted Google college visit for Hopkins. I'm hearing from Jung {}, who works on PageRank. When he first came to Google, he had been working on NLP at JHU. He spent the summer at AT&T Research.
Other dude did ugrad and masters at JHU, then Ph. D. at Stanford. The first couple of years he thought he'd do AI; then the next few years he thought he'd do theoretical computer science. “You end up with a Ph. D., and you've written this thing… but who really wants to read it? I want to do something that matters.” He did research at the DEC research lab, which was fairly practical, but HQ didn't want to hear their ideas. He worked at AMD on the Opteron (!), but he didn't like the five-year cycle of CPU design before he sees results. “Within six months, I can see that things that I'm working on are affecting people.” “I ended up at Google because my boss at Digital was at Google.” “Some people do well at being the world expert at three-level caching structures,” but he likes doing everything. Projects at Google: He started working on hardware. “We're sort of more behind-the-scenes - let's put together the next filesystem, or the locking system - jobs that are spawning thousands of CPUs, and the've gotta cooperate.” Bandwidth's flow across the racks matters, and the liability of redundancy. Real-world scalability.
Josh started last May - first job out of college (MIT). Five-year masters and ugrad. Research focused on optimization and modeling, and that “doesn't get used terribly much at Google, but I was hired more as a general good computer scientist.” They don't pigeonhole you [ed: they might pigeonrank you]; he's working on consumer resource management tool in Python (yay!), and had never worked in Python before.
Badaji (?) - joined in June. Fresh graduate from USC with masters. Did ugrad in India, working on network filesystems in FreeBSD. No one works in FreeBSD at Google. (”That sounds interesting,” Google interviewer says.) Masters in sensor networks, but Google doesn't do anything there. He joined Google as a vague decision - they had filesystem work, and “this would be a good opportunity for me to move into what I actually want to do.” Had him work mainly in Java on AdWords bidding system. Decided he wanted to kernels, and now works 80/20 AdWords/Linux kernel group. Plans to move slowly into kernel group. “That's one of the advantages of working at Google - you get to move into what you want to do.”
This is really rough, and I'd apologize, except I'm not sorry.
“We're not allowed to say a lot about what's not on the website,” says [green shirt] when asked about next steps for Google Corp.
Questions:
- Going public: “Internally there hasn't been any change - basically as software engineers we are still behaving the same way [audience chuckle] - we just do the coding and don't worry about the investors at all.
- “In terms of the culture, I think a big reason people want to come work at Google is the culture at Google. It's a very employee-centric company. The question I was getting constantly was, How are things going to change?…. [Nothing changed.]” “There are more projects now!” chimes in the Indian guy.
- Do you engeineer your own data storage stuff? “Sometimes the real commodity stuff is not quite as reliable as some higher-end stuff, and one thing we think we're very good at is building reliability on top with software. That helps us keep our costs down, by relying on the commodity nature. We do put together our own boxes - we don't go to Dell or Sun and buy boxes.”
- Name a specific problem. Green guy: “We always want to make our search quality better. We evaluate our own stuff…. Some people have some ideas, some of that's just waiting for next generation stuff to roll out. They talk about doing good search being an AI-complete problem - never really heard the term before going to Google, but the idea is that if you solve search perfectly, you've solved all of AI.”
- What's Google looking for? Recruiter: “I mean that's a really general question. Overall we look for people who have demonstrated very execellent academic skills - that's the first thing we look at. We recruit at the top schools, we look at GPA…. We look for people with an entrepreneurial spirit. We hire people that are smarter. If we continued to hire people just like ourselves or people we thought we could teach something to, we […]. We're going to look at the letters from their professors, …, we will look at internship experience. Breadth of knowledge over depth of knowledge, at least for people two or three years out of school.”
- Guy asks, Is Google going to do “real NLP” or just stay information retrieval techniques? ZH guy, “Typical NLP tasks like questioning and answering… We also need to consider the cost and the efficiency. If we increase computer time by 100, will the user's results be 100 times better? … Anyone with this background is real [?].” Green guy, “We'll use more and more of it as we figure out how to make it cost-effective.” (Questioner: Are there are a lot of people working on that question?) Indian guy: “As you must also be knowing, Applied Semantics was taken over by Google, so we have more NLP people now.”
- Who do you want? Recruiter: “One thing Google does that's different from other companies is we don't preallocate. … You go through the interview process”, and they decide where to put you a few weeks before you start. “20% of your time as an engineer can be spent working on anything you like, with your manager's approval.”
- What percentage of the applicants are offered positions? Recruiter: “Considering the amount we interview, and our aggressive goals, it's really hard to say right now. What I can say is that we are hiring a lot of engineers in 2005. I can't give out those numbers, but you would all be stunned. We feel it would be better to have no one in a position than have a warm body.” … “One non-technical phone screen, two technical phone-screens, a committee reviews feedback from those screens,” then they decide this and that and one or two days of interview, and then founders “We do hiring on a consensus basis.” Asian: “The hiring process of Google is much more efficient than any research lab.”
- New York office - what do they do there? Recruiter: Mostly engineers. Indian: “Some of the projects are partially done in Northview, Zurich - we do videoconferencing and it's not like one project is at one center and it's not restricted like that. You could be in New York and still participate in a project that's [somewhere else].”
- Masson: “Do you feel that what you do is somehow recognizable in the product [that comes out]?” No one really answers. Green guy: “Things are still done by fairly small teams - if any of those people weren't there, you can tell it wouldn't really have gotten done.”
- “How do you feel about your coworkers?”: Green guy: “All code has to go through code review. It's a very good group of people.” Asian: “Very cooperative, very helpful.” Recruiter: “We hire people who want to be team players. Anyone who wants to be an individual player won't end up in an engineering role.” Indian: “Right now I have only cubemate and all he does is juggle some balls and doesn't distract me.” Green guy: “They keep us tightly-packed in the office, and it's good because everyone's within walking distance. And so, communication that really needed to have happened happens.”
- “Typical day?”: “We dont' have fixed hours or anything - I happen to be a morning person. My cubemate wakes up at twelve in the afternoon; he plays poker at night. I've done some coding, I eat lunch, play some ping pong. By about five-thirty or so, I go back home, watch a movie or so. You don't really have to work more than twelve hours [per day] or something.”
- Biggest challenge? “At the company, or us?” Green guy: “Living up to the expectations.” Indian guy on right: All the expectations and rumors from Slashdot.
- GLAT weight? Recruiter: “I'm somewhat familiar with this. It depends on the candidate - if we saw a resume that was outstanding, and he didn't do well on the quiz, then we'd consider him. If he was borderline,
- “Are there right answers to the silly questions on the quiz?” “I don't really know,” says recruiter.
- There are some cases where the system will expand it beyond ten words behind your back - beyond the ten-word limit, through stemming, says Green Guy. It becomes a cost issue, we want to keep latency down.
- Are you guys planning on making a browser? [Everyone looks around. Recruiter looks stern.]
- For Green Guy: “Do you feel that your Ph.D. has really helped you in your career over time?” “Definitely. I can tell you I have filed for four patents in my two years at Google, and it's really thanks to Johns Hopkins to give me the background. But you have to use them to solve real problems, not to solve some mathematical things.”
- To what extent does Google use open-source: [Mumble, mumble.] MySQL, Perl, Linux. This guy wants Google to say something about being an open-source company, in the ESR value-adding model. No one on the team has anything relevant to say.
- Code review: Revision control, high-level and low-level feedback - using good sorting algorithms, giving variables good names - usually there are some iterations back and forth. For some very important code, there's a group of people […].
- For you guys who don't have Ph.D.s, do you wish you had one? “No” from Indian guy. “I would like to go back [for interest],” says red-shirt.
The Green Guy was so-named because he was wearing a green shirt and he talked a lot. Feel free to leave comments
Raffle, then T-shirts. Raffle winners: Alak Mehta wins Google fleece jacket. Google beach towel: Ayse Subuntu, but wasn't here. Devin from Entertainers' Club won it. Andy won the other towel. [?] won a pack of pens (sure, why not raffle it?). Rajesh won a lava lamp.
“We're going to distribute T-shirts now. If you could just take one, that would be lovely.”