Programming guru

It's not so hard to find these guys.
A lot are ex-telco, so, 110k is a step up.
Low level network programming and concurrency. Guys who like C.
The finance bit is hard to find. But rarely needed.
 
Quote from Rationalize:

It's not so hard to find these guys.
A lot are ex-telco, so, 110k is a step up.
Low level network programming and concurrency. Guys who like C.
The finance bit is hard to find. But rarely needed.
More...

Not really. Try going that low in frequency and it is not telco - this is a network stack. You likely NEED a Linux here, and you NEED to possibly tweak the kernel. The timings he wants are a LOT more brutal than anything you even have in telco, where you basically have a sound hardware subsystem "saving your ass" timing wise. I did video playback some time ago and we were happy with the 1ms high resolution timer Microsoft gave us. That is 1ms - about 40 times as coarse as the requirement here. Any audio playback was a joke - you add anti jitter buffers anyway and let the hardware more or less take care of things. Nice - a 0.05 to 0.1 second anti jitter delay is not something anyone ever realizes. Except when you need a 0.028ms accuracy to beat.

When we take into account this has to run all the way down to the network stack, you have to start undoing a lot of optimizations. For example network cards - the "latest optimization" is not to interrupt for every packet, but let them accumulate in a hardware buffer. The small delay means a LOT less interrupts. Nice - until you are suddenly trying to hit this type of latency and need a callback NOW.

You really need people that did switches or something - and then know the operating system in question good enough to tune it.
 
Quote from Rationalize:


110K - 200K is about market, tragically.
More...

That's heading towards 1%er income so I'm not sure "tragic" is exactly the right descriptor, but I agree with the general sentiment, at $150k+ you should be able to find decent talent, assuming you have a good story to offer them.

I had a junior (less < 2 years experience) guy implement a zero-copy kernel bypass solution in less than 2 months. It wasn't perfect, and there was certainly room for some improvement, but we're not exactly talking about sequencing the human genome with an abacus and duct tape level of difficulty here. Not in 2013.
 
Yeah, yeah. Turn off Nagle's. Think about FPGAs. Yawn.. Long way from trading.

Tragic, because this is properly hard work. Easier things to do for 110 in my experience.

p.s. Juniors don't deserve duct tape :p
 
Funny how everybody on this website is talking about chump change, then talks about trades and classified ads in the order of hundreds of dollars. Then when someone posts a pretty legit looking offer that none of you unqualified people have a chance in hell of getting, all of a sudden 100k-200k is "peanuts." I don't believe you for a second.

There's a good chance though, that you are unemployed living in fantasy land.

Quote from NetTecture:

I did not ask for the job. Among any other things the locality is a no-go for me to start with. Not even US based and no interest to come permanently live in a country that treats the largest percentage of people living there as criminals ever.

My last offer was the equivalent of - hm - 180.000 for data warehouse work and that is where people are available. This - seriously - goes into the territory where you are going to have to HUNT for people. Kernel driver optimization. Not my area of interest or expertise. Seriously, for this type of work you likely talk to 100 people maximum worldwide that really know those stuff.

You obviously must not pay a million, but really, your offer makes no sense per year.

But lalala funny wages wont get you a magical specialist and have him stay the moment he realizes how you screw him.
More...
 
My view is that this kind of work pays, but not significantly better than more fun work, using the same skill set. I have been there, and done this, and don't plan do it again. There is no upside for guys in this role. Just good experience for the young, naive, bright & motivated. There are plenty of these guys around, at 110-200. Matches what the banks pay almost exactly.
 
I can't PM you, but here's a thought:


Are you co-located? Connecting directly to CME, ICE, etc, etc. is not equivalent to co-located within a few meters of the exchange matching engine. You will shave a few micros doing that. But it can get costly, as ICE matching is probably located in GA or NY. You have to be aware that it's not the issue of programming in C++ (ie. software acceleration) to reduce latency but you will require hardware acceleration if your talking in the 'µs' realm (co-location at Aurora for CME products and/or financial FPGA)


I have experience with and have developed custom front-ends and strategies for: C++/JAVA; TT FIX Adapter, TT API, CQG API, CTS T4 API, Firmsoft/Clearport for trade capture and real-time P&L. For a few prop shops, a hedge fund and currently working on my technology consulting business.


Depending on scale and AUM$; If you want someone to do this... you're looking at anywhere between $250k-$375k a year.. anything else would require the lesser salary and a % of P&L.

You can also try to outsource to reduce costs but then it will be difficult to maintain IP...



Quote from garachen:

Mid to high frequency trading company looking for a programmer who can push latency below 20 micros.

We have several functioning strategies at our current speed and have spent the last several months connecting to new exchanges. We connect directly to CME, ICE, CFE, eSpeed, BrokerTec, Arca, Bats. Looking for a Linux/C/C++ guru who can lower our latency the last bit and who has a thorough understanding of multicast fees.

Full time. California.

Salary range: 110 - 200K depending on how good you are

PM me if interested.
More...
 
From the majority of comments I've seen I think it's safe to say that posting this on ET was perhaps not my brightest idea. In my defense, this was always considered a long shot.

Luckily, my other leads panned out so I'm content.

Wanted to clarify a few things though - put some information out there.

Base salary of over 200K for a programmer in this industry is too high. There are plenty of excellent people available at this price to do what I'm asking. Except in rare circumstances if you are paying more than this you have overpaid. If your programmer is so essential to what you are doing you should consider a much lower base/ high bonus or equity.

20 micros is fast. 10 is wicked fast and 5 is insanely fast. FPGA's are being phased out. Microwaves are getting cheaper. Sandybridge is awesome.

I think it's safe to assume that anyone spending a few million a year on tech knows how to co-locate and where the ICE matching engine is located. If you accidentally put your server in Georgia you ought to move it.


Anyway, as customary, the dialogue here has been simply inspiring. Once again the anonymous horde has reared it's collective wisdom to strike down the silly job poster. I humbly apologize as I slink back into the real world.
 
Well, I was not aware where ICE matching was located. I've worked with marketmakers who used WebICE's ICEMaker to send quotes to the exchange directly from their trading work stations in NY/CT. Pretty low tech, but got the job done. Never worked with ICE co-location and thus unaware of location. Sorry if you found my comments as striking you down; I believe I was the only poster not shooting you down or being insulting.

Good luck :D




Quote from garachen:

From the majority of comments I've seen I think it's safe to say that posting this on ET was perhaps not my brightest idea. In my defense, this was always considered a long shot.

Luckily, my other leads panned out so I'm content.

Wanted to clarify a few things though - put some information out there.

Base salary of over 200K for a programmer in this industry is too high. There are plenty of excellent people available at this price to do what I'm asking. Except in rare circumstances if you are paying more than this you have overpaid. If your programmer is so essential to what you are doing you should consider a much lower base/ high bonus or equity.

20 micros is fast. 10 is wicked fast and 5 is insanely fast. FPGA's are being phased out. Microwaves are getting cheaper. Sandybridge is awesome.

I think it's safe to assume that anyone spending a few million a year on tech knows how to co-locate and where the ICE matching engine is located. If you accidentally put your server in Georgia you ought to move it.


Anyway, as customary, the dialogue here has been simply inspiring. Once again the anonymous horde has reared it's collective wisdom to strike down the silly job poster. I humbly apologize as I slink back into the real world.
More...
 
yeah man, making 6 figures working in financial industry is so naive, only fit for a youngster straight out of college. UH HUH. It's much better to dream about making millions while gambling on a computer screen in the squalor of a rented apartment. Yeah that's the elite trader way, go for broke!

Quote from Rationalize:

My view is that this kind of work pays, but not significantly better than more fun work, using the same skill set. I have been there, and done this, and don't plan do it again. There is no upside for guys in this role. Just good experience for the young, naive, bright & motivated. There are plenty of these guys around, at 110-200. Matches what the banks pay almost exactly.
More...
 
You must log in or register to reply here.
Back
Top