At that moment, All I had in my head was a bunch of things and an urge to search for it. However, it was pretty clear that it was not in Singapore. I considered this to be like a high stake bet. In the end, I decided to leave a PhD program that includes a full scholarship and an opportunity to have a one year exchange in CMU. A week later, Harvard Business School sent me an invitation for their Doctoral Business Program with a full scholarship, the top lab equipment, a computer and a promising career. They did not change my mind.
To me, having a PhD means having a bare minimum requirement to join academia and become a Professor. If I fail this bet, it means that I will make myself look terrible in front of my academic supervisors, so they are. However, what I seek is knowledge, not a title. Truly living in that experience is more important to me than the exterior validations. I went all-in for that decision.
Machine learning used to look like a joke to me compared to all of my previous courses in finance and physics. However, the painful part behind the scene is engineering specialized for processing big data. I was pretty confident at the beginning given my three years learning computer science with a university curriculum in high school. It was a boring experience when most of my time was spent debugging code, so I returned to pure math and did graphic design as a hobby. Interestingly, drawing on computers turned me into the top candidate of several student projects that linked to a startup world.
In the first four months, I became a resident of the internet. I took courses, learned to program on an open free education platform in a more serious way and tried to build some toys on AWS. At that moment, MOOCs were just taking off. Machine Learning was more like an exciting course for a small group of students to try something new with MATLAB. I even purchased a software license to do proper research without any idea about open source projects. Python looked to me as a nice and friend-ly programing language to learn but was not so attractive.
A competition on the internet for computer vision opened their annual challenge. They provided data and sponsored a computing power to run a result for 15 teams worldwide, so I signed up. Working on deep learning at that moment was like a privilege because you would need a GPU to run a program in a few days. Last time I tried to work on AWS with a CPU and did not terminate a machine. I received a bill that cost up to more than a hundred dollars which was way too expensive for a test of deployment.
Luckily, I was chosen and had a few K80s - the top cutting edge technology of Nvidia. Until then, I just realized that there was a giant gap of skills that I needed to fill. Working remotely on a machine is much different from working on a computer. There was no graphic user interface or a pointer (a mouse), but a black screen of a terminal. That was when I realized that I had to ask around to find someone knowing engineering better than me. After a few days, I was able to get access and work with a command line. It felt like back to the dawn of the operating system with MSDos.
Quickly then, several problems came to me. I kept losing data every time my local machine locked off, but I was able to fix it. However, there was a bigger issue - I did not know how to configure GPUs. One change on the source code also took me lots of time to debug given a laggy internet. I was totally knocked out when realizing that to master this game, I needed to know the ecosystem of open source written by Python. That was a serious crisis when there was no learning community, poor documentation and guidelines to this roundabout game. That required lots of hands-on experience on technology for big data to refine knowledge. It was too much for someone with a theoretical background and to avoid coding like me.
That was when I decided to join the industry for help. At least, they could help me upgrade my technical skill set. I tried to apply to companies and organizations that worked on related technologies. It started with my city first, then a region, eventually I moved to the US where most of the content was from. It was very underwhelming because there were only a few choices and I kept having rejections. Working on big data needs lots of experience and no one would open a position on deep learning except for big techs and a few hot startups in the Valley.
I then found myself with data science, a job that I read in an article of Harvard Business Review in 2012. At that moment, it was hot in NYC and San Francisco. There were not enough people having the right background to work on this fancy job. There were programs that only hunted people from the academic world to make a transition into industry. It sounded a very great fit to me and that was how I found my path to the Valley.
2015 was a special year for that competition. A team from Microsoft Asia crossed the benchmark of algorithms with 90% accuracy. This was the first time ever a machine exceeded human performance. It marked a chapter of racing between Big Tech. Top thinkers and billionaires debated about the benefit of AI towards mankind. This led to an organization called Open AI as the non profit led by some most promising tech executives in the Valley. Google turned their X project into Google Brain.
At that moment, I was just able to upgrade myself with some Ruby, build a blog on Github, and write some first thoughts about AI in Vietnam with no travel plan.