nVidia CUDA Spotlight Week in Review
GPUs in the Big Apple
This week's spotlight is on Andrew "Shep" Sheppard. Shep is a financial consultant with extensive experience in quantitative financial analysis and trading-desk software. Most recently, he was chief technology officer and chief quantitative analyst at a New York multi-strategy hedge fund. A CUDA developer and published author (with technical publisher O'Reilly), Shep entered finance after conducting scientific research at Oxford University, Caltech's Jet Propulsion Lab and the Berkeley Space Sciences Lab, where he worked on earth and planetary remote sensing probes.
NVIDIA: Shep, tell us about the current landscape in finance and computing.
Shep: Data in finance is exploding. So too is the velocity of the data, by which I mean the speed and direction in which it moves from place to place (for example, prices for similar assets quoted on multiple exchanges or trading venues being pulled into a ticker plant). And there is a pressing need to analyze this data to make it actionable. To meet this deluge of data and analysis, and the push to make everything run in something near real-time, the GPU is seeing wide application.
NVIDIA: Where is the momentum?
Shep: I am seeing the GPU gaining momentum in a number of key areas: pricing of complex assets, such as collateralized debt obligations (CDOs); moving risk from overnight batch processes to real-time; and back testing of strategies for high frequency trading, or HFT as it's known.
NVIDIA: What sorts of problems in finance are GPUs a good match for?
Shep: A surprising proportion of financial problems are what's known as embarrassingly parallel, in the sense that they are very easy to map to parallel technologies, such as multicore and GPU, with spectacular speedups (10X, 100X and beyond). I would turn the question around. In finance, you may be hard pressed to find problems that aren't a good fit for HPC and GPU supercomputing!
NVIDIA: What are you working on now?
Shep: I help people in finance make more money by applying supercomputing technologies. These projects cover a wide area, from real-time risk to HFT. I am also active in the HPC/GPU space generally. I've set up special interest groups and meetups in New York and Boston, and those have been wonderfully successful in a very short time, reflecting a surge in interest in HPC and GPU supercomputing. And I have a couple of books ("GPU Supercomputing in the Cloud" and "Programming GPUs") in the pipeline with the publisher O'Reilly (who is, in my opinion, the best technical publisher on the planet!).
Related Articles:
|
