Entries Tagged 'OT' ↓

We're hiring

Or, we'll hire if we find the right person.

If you live in Israel and are into things like:

  • hardware/processor/programming language/library design
  • compilers, debuggers, profilers, simulators, instrumentation
  • number crunching/optimization
  • parallelism/multi-core/distributed computing

…we might have a bunch of stuff to interest you. Benefits/drawbacks/features:

  • Most projects are intended primarily for internal use – "infrastructure" if you like
  • Lots of custom things, ranging from our own chip to our own distributed build server
  • Plenty of production projects depending on your work
  • Ability to do grand things singlehandedly guaranteed through understaffing
  • Bleeding edge, "first/best" system in several categories
  • Little management, especially if you don't really need any
  • A "Worse is Better" spirit (reasonably pragmatic perfectionists are welcome though)
  • Many long-term projects
  • No downsizing during previous bubble burst
  • Experience in any specific relevant area appreciated but not required
  • Mostly C/C++, Python, and our own languages; if it matters, you can use any other language
  • Mostly Linux, a bit of Windows, Eclipse usage share eclipsed by emacs & vi
  • Nice people
  • Jerusalem

A great place to work, if you ask me. If you're interested, send mail to Yossi.Kreinin@gmail.com

The Internet age/reputation paradox

A person's reputation tends to rise together with the age. The older one is, the more opportunities one had to do notable things, and to meet people who could appreciate those things and tell others about them. So this makes sense.

An online document's reputation also tends to rise together with the age. The older the document, the more documents link to it, and the more documents in turn link to those documents, raising the old document's PageRank. So this makes sense, too.

The paradox is that the older documents are written by the younger people. That is, it is one's younger version that wrote one's older documents. So the documents with the most reputation will tend to be written by people (or more precisely snapshots of people) with the least reputation; one's dumb young stuff may well pop up first in a Google search.

(Not that there aren't any counter-tendencies to cancel this effect at times; my old anxious, moronic report of an imaginary bug in ALL CAPS no longer shows up in my egosearches. So no, I'm not bitter. In fact, Google loves me more than I deserve – for instance, my review of Extreme Programming Explained has appeared in search results right after the Amazon entry for the book ever since I published it, and I've only skimmed through the thing. The only thing that bothers me in the SEO department is that the search for "C++ FQA" gets corrected to "C++ FAQ" – didn't expect that once the query got past the spell check barrier. I hope my collegue's riskily named DreamPie project will not experience a similar setback.)

Applied mathematics in business consulting

As a part of my continuous moral degradation and the resulting increasing alignment with the forces of Evil, I'm sharing an apartment with a gal who used to work in HR assessment. She recently got me acquainted to a friend of hers, BC, who works as a business consultant (names have been changed to protect the guilty).

BC's primary educational background is in applied mathematics. Having put the math they teach in CS departments to relatively few uses as a working programmer, I asked her about the uses of applied mathematics in business consulting. BC cited the following two examples.

The first example involves compensation and its dependence on key performance indicators, affectionately known as KPI and estimated by HR assessors. One way of looking at this dependence is to consider how it affects compensation over time as an employee's competence increases.

A psychological discussion is then possible on the relative merits of the different graphs plotting the compensation functions f(KPI). If f is linear (has a constant derivative), we make people struggle equally hard at every step. If f's derivative increases over time (for instance, when f is exponential), we make elevation hard at first and then increasingly easy. If f's derivative decreases over time (for example, if f is logarithmic), we make the last mile the hardest. And so on.

Through a psychological discussion of this sort, someone in the consulting company decided that what was really needed in some case or other was an S-shaped curve. The problem was that you couldn't just draw an S-shaped curve – the plotting had to be done in Excel according to a formula; an S-shaped curve which just blithely goes through arbitrary points doesn't cut it when you deliver a Compensation Model. But how do you make a formula to go up slowly, than fast, than slowly again? Exponents don't work. Logarithms don't work. A sine does look like an S, but it's a wrong kind of S, somehow. What to do?An S-shaped curve

Enter BC with 6 years of math studies under her belt. A compact yet impressive formula is spelled out, and – presto! – Excel renders an S-shaped curve. (I guess she used the sigmoid function but I didn't check.) The formula brought delight to management and fame to BC, and compensation payments issued according to its verdict keep adding up to scary numbers (BC's agency works with some really big companies).

The second example involves the compensation of managers. Naturally, a good manager near the bottom is worth less to the firm than a bad manager near the top, and therefore the compensation function should now depend on the manager's level in the hierarchy as well as his KPI (or better). Equally naturally, the numbers coming out of the compensation spreadsheet will under no circumstances arise through an externally conducted study of their psychological implications or any similarly unpredictable device. The numbers will result from nothing but the deep understanding of the organization possessed by the top management.

The development process of the managerial compensation function is thus complementary to that of the employee compensation function. Instead of producing numbers from a beautiful spreadsheet, what is needed here is to produce a beautiful spreadsheet from the numbers specified by the top management. The spreadsheet then promptly generates back these exact numbers from the input parameters.

The purpose of the spreadsheet is to relieve the top managers from the need to justify the numbers to their underlings. In order to guarantee that they are relieved from this need, the formula should not contain terms such as 1/(level^2), which could raise questions such as why not use 1/level, why not use 1/log(level) and other questions along these lines. Rather, the formula should contain terms which could raise no questions at all simply due to their size and shape.

BC faced this problem at an early stage of her career, and despite the natural stress, came up with an interesting Compensation Model, its key term being e raised to the power of something unspeakably grand, combining the trademark Gaussian look and feel with an obvious ability to deter the skeptics from asking questions. The only problem with that term was the very source of its utility, namely, the fact that it evaluated to 0 for all values of KPI and hierarchy level.

The deadline being close, BC told the manager of the consulting project in question about the status of her work and expressed her doubts regarding the delivery of the Computational Model. The manager told her that she just doesn't get it, does she, it's great, the right numbers come out and that's all there is to it and we should send it right away. And so they did, to everyone's complete satisfaction.

Her command of applied mathematics aside, BC is generally quite powerful.

For instance, she once got invited to consult some government agency about a project of theirs, while being on vacation and without it being explained to her that she was about to attend a formal meeting with the whole team. In order to maintain the reputation of the guy who somewhat clumsily brought her in, she had to improvise.

The project, worthy of a government agency, was related to some sort of war on corruption, the unusual thing being that they wanted to fight the corruption of other governments. Their weapon of choice was the training of representatives of another government, financed by the other government, in their supposedly superior methods of governance. While the general concept was impressive on many levels, the details were unclear.

BC had to speak, and she spoke based on a principle appearing in a book by some McKinsey alumni (she didn't tell its name nor generally recommended it): whatever you tell people, it should contain 3 main points. Possibly 4. But preferably 3. More is overwhelming and less is boring. So she said: "At its core, your project is about teaching people. It is therefore essential to clearly understand three things:

  • Whom you're teaching,
  • What you're teaching them,
  • And how you're teaching it."

And they started writing it down.

I asked BC whether there was some way to unleash her on the company employing me so that she grinds a few bullet points into them (a handsome Compensation Model being as good a place to start as any). She said something to the effect of "it only works on the weak-minded"; it was apparent, she said, that the government agency in question had little previous exposure to consulting.

BC says she (still) believes that business consulting is meaningful and valuable, which sounds paradoxically at this point. But, looked at from another angle, it really isn't. Don't view her stories as ones undermining the credibility of business consulting but rather as ones building her own credibility as a person aware of the actual meaning of things and willing to sincerely share that understanding (how many people would instead say that they Developed Cutting-Edge Compensation Models?) If she says there's meaning to it, perhaps there is.

Off topic

  1. To comment, you no longer need to register, just type "y" to confirm you're a human. Thanks to Aristotle Pagaltzis for pointing out that the previous arrangement sucked.
  2. I've started another blog, mostly hosting images. For example:

I originally intended to have one blog for everything, but since you've probably subscribed for the technobabble, I'll reserve the channel for that.