DE Shaw & Company takes its very smartest people (and a non-trivial portion of their profits) and puts them to work at DE Shaw Research, where they work on protein folding and other computational biology that could help cure cancer and HIV. I think that's a pretty worthwhile use of talent :-D
From what I was told, David Shaw is spending most of his time on the research group these days.
They ostensibly don't consider themselves a finance firm - they consider themselves a tech firm that will use their skills wherever they can to increase efficiency and make money.
The problem is that DE Shaw Research made a loosing bet recently when they decided to design custom silicon to solve the protein folding problem while everyone else kept using commodity hardware and worked on improving the software.
I saw a guy I know at a wedding a few weeks ago who works at Shaw Research, and I asked him about this. He didn't give explicit details on custom vs. commodity performance, but he did say they had machines up and running and giving novel results.
I doubt it is fair to say they are ignoring software improvements. I saw Shaw talk about the machine architecture and he had a lot to say about balancing programmability for later software improvements vs. specialized computational resources. Also, it was his algorithmic improvements, the "neutral territory" methods, that were inspiration for the machine. Are NT methods still state of the art for molecular simulation?
How do you know that this is a losing bet? It's still pretty early days. And aren't they hedged with their work on Desmond, which is MD software for commodity clusters?
Custom hardware for chemical physics has been tried multiple times before since the late 70s. I’ve heard most companies only found financial success through bloated government contracts and the custom hardware only provided limited advantages for researchers. In all cases these advantages died out quickly with advances in commodity hardware.
Anton is impressive in that it can provide millisecond long trajectories of protein/solvent systems, but single trajectories are of limited utility. You need many (1000s) of such trajectories for statistical analysis of the molecular system. Further, there already exist several clever methods that leverage chemical statistical mechanics to provide the same analysis without the need for single long trajectories. As an example, see Pande’s work developing Markov models of protein folding using millions of short trajectories between metastable states. This method has already provided a complete statistical analysis of protein folding for proteins that fold on times scales an order of magnitude beyond what Anton can simulate.
As for Desmond, there already exists a plethora of free MD programs (MMTK, LAMMPS, NAMD, CHARMM, Gromacs, and many more). Many of these, especially Gromacs, have already been highly optimized for a range of hardware and I wouldn’t expect Desmond to surpass these free codes by a margin worth dropping dollars.
Personally, I still have high hopes for DE Shaw Research, I just don’t see how their current offerings will turn a significant profit or greatly advance science. I’d love to be proven wrong, and I’m sure they’ll have plenty of additional novel future projects, some of which could be paradigm-shift-changing for chemical physics and molecular biology. My guess is that such advances won’t come from their hardware geniuses, but instead from their math/physics geniuses that will develop new statistical mechanics methods to bend & contract in silico time.
i'm not convinced they're smarter than anyone else. i'd rather invest in renaissance or citadel...
my coworkers at morgan were all physics or cs phds. and a consulting expert got a nobel prize while i was there. amusingly, i interviewed at google around the same time, and some of the engineers were condescending about the quality and education of my coworkers.
Renaissance over Citadel. If I'm not mistaken, Renaissance is purely automated trading. No traders or trading floor in the conventional sense. Plus they weathered the fallout of 08 much better than Citadel.
What's sad is that these smart people, including ones I've met while working as a SysAdmin for a Prop Trading firm often don't have access to the capital to start their own shops. This is due to the incestuous nature of the Finance world, where it's a lot more about who you know than what you know.
Sadly, they're often in the employ of third-rate CEOs, who always get a cut off the top, and pay the producers a mere fraction of what they made for the firm and their clients.
The Medallion Fund has its own internal trading desk, staffed by approximately 20 traders, and trades from Monday opening bell in Australia through Friday closing bell in the US.
it was just an offhand remark. "if you came to google, you'd get to work with mostly phds, instead of who you work with now" -- without even asking about the kind of people i worked with.
granted rentech probably has desco beat (at least in terms of recruiting toughness - the only metric I'm really in a position to judge by). But I don't think anyone would deny that desco is part of that top tier too.
FWIW, in my experience the handful of citadel programmers I've interacted with aren't quite in the same league as shaw/rentech/getco/jane st/etc. I have no idea how that reflects on their returns though ;-)
Peter J. Weinberger (W in AWK) worked at RenTec before moving to Google.
I would not invest in anything from them however. Medallion has been closed for new money for many years now and all other offerings performed poorly to say it politely.
I assume their business model is roughly : to implement algorithms that find near-arbitrage opportunities by sifting through vast amounts of market data, then use those to multiply investment funds. This sounds like the perfect domain for a startup.. where a nimble small team of quant/developers would have huge advantages, by being faster to rollout.
This is almost a pure software business, so why aren't these companies more like startups? Is it because of the expense of getting fine grained and low latency market data [eg. exchange colocation]? Or is it due to the contacts needed in the biz to get the 'investment' funds to trade with? Or is it because only large financial entities have access to these risky speculative trades?
Side note : I see Jane Street Capital use Ocaml, and some other quants use KDB/Q for implementing these kinds of algorithms, so it seems innovative languages give leverage here. I was thinking Node.js + a js BTree api to access streaming data would be a nice dev environment.
Two major difference of a trading firm from a software firm:
1) Large amount of trading capital is needed to make a sizable profit.
2) Math, finance and trading skills.
A trading firm needs above 2 items in addition to technical skills to succeed. A team of good people with all 3 above items (capital, trading, tech) have a good chance to succeed.
In fact, Citadel (one of the largest quant hedge funds) was started by one person (Ken Griffin) when he was a undergraduate from a Harvard dormitory. It is pretty much a startup success story.
There is a major culture difference: trading is the key activity; coding is only secondary. This may explain why trading firms usually have a typical wall-street tough culture, and don't feel like a typical silicon-valley startup.
Not so different as you might think. After all, a large amount of capital is needed to build out a datacentre. Doesn't mean that Google is the only company that can make a profit on the Internet.
Not really on the same scale. If you're a hedge fund, the most you can make in a year is some small percentage of your total assets under management, so you need to persuade people to give you billions (at least hundreds of millions) before you can start raking in the kind of dough that pays salaries.
Nice link.. seems like HTG have taken the first step towards a pluggable infrastructure, where the plugin is the quant/programmer + their algorithm implementation.
ps. I never really understood the difference - Hedge Funds seem to be more about speculating via leverage than 'hedging'? [using their own proprietary algorithms to do that]
How much staff is really really needed for these quant shops? After an algorithm is developed isn't it just rinse/repeat? Honest question, as I'm not a finance guy.
Just like a technical firm needs to iterate on the products and infrastructures. A trading firm needs to iterate on the algorithms and infrastructures to stay competitive.
Market changes constantly. Many trading algorithms are essentially market-data driven and need to adapt to the market or lose.
One thing I always found odd about DE Shaw is how casual the work environment is. It's like google, but in Manhattan. The problem is NY isn't california and quant finance isn't all software. There's a certain pace and overt competitiveness to Manhattan that enforces a discipline and a work ethic onto people. This works quite well in the Darwinian world of finance.
But, from what I saw of various parts of DE Shaw's operations there was somewhat a lack of urgency to some fairly critical areas.
A good firing spree every once in a while seems to work wonders in putting the urgency back into the workforce.
Google has an office in Manhattan, 15th street and 8th avenue if I'm notmistaken. I haven't personally been to any other google office, but it doesn't look different than the pictures of any of them; All the perks, sleeping pods and cafeteria stuff seems to be there.
> Intel announced it will reduce its workforce by 4,000 workers (5%), mostly through attrition or voluntary separation programs.
I joined Intel shortly after that article was written (2002). For much of my tenure (3 years), Intel had a US hiring freeze (except through acquisition, how I got hired) and aimed to reduce its headcount primary through attrition, not layoffs. I felt like Intel treated its employees pretty well.
Jack Welch always got a lot of flack from people about "differentiation" (GE's program of cutting the bottom-performing 10%).
I think the key (and I don't have a lot of experience with folks who work at GE, so I can't speak to it's validity) is that it works if deployed as part of an overall system.
That system is that you reward the top performing earners heavilly (with bonuses and promotions), and remove the bottom 10% annually. In theory, people are constantly being kept abreast of their performance, and given opportunities to improve; so that those in the bottom 10% shouldn't be surprised when the end of the year comes and they are laid off.
Welch defends the practice by talking about the alternate scenario where people are kept on staff for years who aren't performing well, have never been told that is the case, and are suddenly surprised when one day the company needs to have layoffs and are terminated. He calls it "false kindness".
I've always like the part of Differentiation where you let people know where they stand. I've thought it would be nice to work somewhere that had a better model for performance reviews (where I work presently, they are basically a checkbox).
I'd love to hear from someone at GE as to whether the system actually works the way Jack Welch describes it.
Assuming you can actually figure out who the bottom 10% are. Also assuming that the bottom 10% is a constant and not something that literally changes daily.
Everywhere I've ever worked, people's performance isn't that wishy-washy. There has always been a "normalization" of performance that becomes apparent based on people's fit within the organization and abilities.
Now, I agree that quantifying that is non-trivial. I've always been interested in places where they actually do that (and wonder how accurate it is), but I wouldn't be surprised to find that people who find themselves in that lowest 10% don't find themselves there on a whim.
