It’s been a full year here at NVIDIA and now seems as good a time as any to reflect. The developments and pace of the deep learning research field has been incredible and is well-characterized by the NVIDIA deep learning blog. Let’s take a little time to delve into the perhaps underappreciated world of Solution Architects (SAs), the group with whom I caucus. dgx

For a high level perspective, SA VP Marc Hamilton’s blog illustrates the challenges of SA management. There are numerous and often underpublicized instances of SAs working to provide technical assistance. One exemplar is long-time NVIDIA veteran Bob Crovella on Stack Overflow, who tirelessly answers CUDA related questions. Others have even written the book. A quick poll of the NVIDIA SAs in North America indicate ~27.4% have PhDs and 100% of them salt of the earth. Namely, SAs are externally facing technical resources often with experience from a previous field or vertical, e.g. medical, oil and gas, media and entertainment, high performance computing, etc. They correspond with an external technical resource on proof of concepts and technological integrations to ensure a satisfying implementation. The shared knowledge base among SAs (with appropriate Chinese walls) is a critical advantage. There are SAs at many tech firms in the Valley including Amazon, Salesforce, Cisco, Cloudera, Verizon, HP, IBM, etc. however NVIDIA SAs often have a special bent towards deep learning. SAs are often on-site and in the field (at least one formerly in the British intelligence).

Recently, I had the opportunity to be in the field for a 79 day European rotation. The highlights of the assignment were giving an invited talk at the University Hospital of Bern (Inselspital), assisting researchers at ETH Zürich on extending the 1 bit SGD training algorithm to activations, and attending the Neuroinformation Processing Systems conference (an excellent summary here: https://blog.ought.com/nips-2016-875bb8fadb8c#.hj4jzpktv) all while continuing support of existing technical engagements. It was truly a once in a lifetime opportunity to deepen DL technical expertise as well as forge contacts across continents. While I was the first in this program, it’s undoubtedly a marquee advantage to NVIDIA SAs.

2016 has been an incredible year of firsts, the first Deep Learning Institute (where I was first to work on it), the first DL supercomputer in a box, and behind the first ranked CEO in America (according to Harvard).

Let’s hope 2017 brings as much excitement and advances as 2016.

dgx

If you think this is good, let's pay attention to what's next.