"Levi has deep expertise in Cuda engineering. I would highly recommend Levi for Cuda related questions.
We met for 30+ minutes. I took 3 pages of Notes. Levi explained to me about:
* thread warps vs thread blocks
* when __syncThreads are used
* volatile tag + memory barriers
* memory coalescing (why threads read memory in "gridSize increments")
* cuda profiling issues (nvvp, visual profiler), occupancy
* dynamic parallelism + existing tools (cub + thrust)
Basically, if you find yourself reading sections of the "Cuda C Programming" 3+ times and still not understanding it, just hire Levi -- he'll clarify all your questions.
Lastly, Levi's humbly also very clear about issues he doesn't consider himself an expert in. I threw 2-weeks worth of stored up Cuda questions at Levi. Levi answered 98% of them in authortative detail, and was upfront in stating that the other 2% was out of his domain of expertise."