When I look back upon my carreer so far, I feel very fortunate to have had the opportunity to work at very interesting places, and to meet some very interesting people. I keep very good friends at all my former places of employment (well, the ones that still exist anyway).
1988-1996: NASA/JPL
That was my first job, straight from France, and I just couldn't believe it. Working at a NASA Lab, in California!
The interesting thing about JPL is that if you squint hard enough at my business card right here (yes, I have smeared the phone numbers because they probably belong to someone else now!) you will notice that the only mention of NASA is in the email address. That is because, if I recall, JPL is managed by Caltech, but on land and with equipment that belongs to NASA. The employees, on the other hand, are Caltech employees. But very often the NASA folks visit to see what the peeps in Pasadena are doing with their stuff.
JPL is a really great place to work, it especially was for me since I was brand new in the country and I had a tremendous amount of learning to do, in terms of language, culture, etc. Everybody there was always so helpful, I still remember bringing lunch conversations to a standstill to have an idiom explained.
The pace of work was steady but not too fast, which was good too. If I had started with Disney I would probably have burned out quickly. JPL is geek heaven - plenty of very, very smart people, and enough slack in the schedule that it is possible to engage in side studies when something interesting comes up.
At JPL I started doing graphics and general purpose programming, just helping the scientists prepare their presentation graphs, that sort of thing. When the sysadmin in our group (Oceanography) left, I took over for a few years, and then later, since I had never really stopped doing graphics, I went back to that, but the higher end kind (videos, presentation graphics for litos, etc). I was always very interested in computer graphics, and I was always on the hunt for computers that were just getting installed and offered to "test" them.
It is only suitable then that my longest lasting contribution from my
time at JPL is this picture. It shows a map of
TOPEX/POSEIDON
data mapped onto a globe, and of course some topography on the land
masses to give it a bit more appeal. There are actually 3 such images,
all slightly different (has to do with the satellite ground track), some
day I will extract them from their digital sarcophagus if I can find
them.
With this image I had to reinvent what I did not know existed in the movie industry: compositing. It's actually 3 separate images, a background plate, the satellite (extracted from another litho which can be seen here, and matted - another technique I did not know existed) and the globe + cone. The original was a humongous 6400x8000, which even now is considered very high resolution, and in 1995 it was really, really hard to do. The JPL image lab printed some Ektachromes of that which were to die for, they looked so amazing. Well, after I had also rediscovered gamma correction, that is!
This image is very obsolete, 8 years old, and yet I still see it places. Here for example, and I even get credit at the bottom of the document, with many thanks to my good friend Kelley. And here, and here, and here, and here, and here (with a link to a nice "hires" version here)... It's really nice to see that it's still around. My JPL legacy!
1996-1999: Disney Online
In early 1996, the web was still relatively new, and Disney was starting to build up their internet presence. Pretty heady times to go work in that field! I had to try.
During the time I was there, the Online division went from about 60 people to over 700, changed names 2-3 times (I can't recall), went through many reorganizations. Compared to JPL it was simply dizzying. I don't think I ever work so hard in my life as when I was at Disney either. But, as Nietzsche once said...
I started as a Webmaster, that is what my business card said, I never bothered to make new ones when I changed jobs. Webmaster was a fine enough job, although I was always drifting back towards programming and that wasn't really something I was supposed to do, since roles at a place like Disney are far more precisely defined than in a more "freestyle" academic environment such as JPL.
So I finally segued back into programming full time, with an interesting assignment: web site traffic analysis
Disney has refined marketing down to an art, and the web site is also a
marketing tool. So it was very important to know what pages people looked
at the most, how they went through the site, etc. In addition, Disney is
composed of a large number of mini-companies called business units (Disney
World, Disneyland, Disney Pictures, etc) and since Disney Online was running
the web site for all business units, they all had very specific
web traffic reporting requirements.
The web site had a lot of traffic. When I left it was approaching 50 million hits a day, which back in those days was enormous. The problem was, it was really difficult to process these logs and prepare all these reports in 24 hours. No commodity machines would do it.
So, most of my contribution at Disney was to understand how to optimize log processing. This took a long time and much desperation, because at the same time I still had to contend with the daily avalanche of data with very inadequate systems. It was a busy time. I finally managed to draft a very optimized log processing architecture, and I was looking into running it on a medium size multiprocessor machine when an astute coworker (thanks, Scott!) suggested that maybe I should consider a Beowulf cluster. That was a very new thing at the time, and it took some additional work because log processing is not easy to decompose in completely independent threads, which is what Beowulf clusters are really best at - the network is the bottleneck - but I finally did it. The sysadmins had a really good time getting eval equipment from everybody and setting it up for me, and we quickly proved that the technology would work, and be far cheaper than a large NUMA machine.
Disney switched to the cluster-based log processing system in early 1999. I believe they used it for quite a while after that. It was quite optimized for what it did, so I can imagine that it would have been hard to replace that. There are now better tools available for web site traffic analysis than brute force log processing, which in many cases has become simply impossible due to sheer volume, so I think Disney has moved away from the homemade cluster now.
1999-2002: The great dot-com roller coaster ride
In June of 1999 I joined a bunch of friends (and former Disney coworkers) at a little dot-com called Checkout.com. These were the halcyon days of the internet age. April 2000, when the bubble started to pop, was still many months away.
At Checkout I continued working on web site reporting, helping the marketers pursue the elusive dream of knowing the customer, but (that was my personal influence) without infringing upon their privacy. Things got more sophisticated, with databases connected to the web site data... I learned a great deal about database-backed web sites at Checkout. Also, worked very hard, not quite as hard as Disney, but far smarter. I had some really bright people working on my team (Hi Li! Hi Chuck!).
We had a great bunch of people there, a fantastic technical team, top to bottom. Unfortunately, the bubble popped, and things started going downhill, slow at first, then accelerating. The company changed names a couple times, reinvented itself, but it was in vain. It is no more, I am sad to say. We gave it all we had, but that wasn't enough.
2002-present: Hollywood!
My latest job includes pretty much a little bit (or a lot) of all I've done before: high-end graphics, very large scale data processing, and database-backed web frameworks with reporting and analysis.
I run the software group at Pacific Title and Art Studio (yes the web site shows some of my PHP handiwork too, together with a very clean design from my friend Amber DeFabio). Pacific Title (or "PacTitle" as it is often called) is a motion picture services company in the heart of Hollywood. So I guess that means that I am working "in Hollywood, for Hollywood".
Pacific Title does a lot of different things, which makes life interesting for the software group because we are here to help integrate everything together. Among other things.
I personally tend to specialize in process automation and database-backed
workflows, with a minor in image interpolation and resampling techniques
because, well, I have to have some fun too (see geek).
There are some highly talented individuals on the team working on digital
dirt and scratch removal (very important, since some of the film we get for
scanning has scratches, and almost all of it will have at least some dirt),
color space manipulation, graphics tools architecture, as well as building
tools for the humdrum but absolutely essential business of moving hundreds
of gigabytes of data around, day in and day out.