Advice for Data Scientists from an AI Business Insider
Now that scheduling meetings is relatively seamless (just CC Amy or Andrew in an email, or Slack them a request), Alex Poon is setting his sights on making sure patients take their pills and follow the treatment programs designated by doctors. Alex is currently the VP of Engineering at AiCure — a company that utilizes image processing as a diagnostic tool, as well as computer vision to check adherence of medication treatment plans.
He is also one of the founders of x.ai, which created artificially intelligent chatbots who act as assistants in scheduling meetings through natural conversation in email and Slack. The decision engine underlying Amy and Andrew’s deterministic responses are so good that some users mistake them for actual people, and send them Christmas presents during the holidays.
Alex’s expertise in AI stems from his education and experience in the field. A bachelor’s degree in electrical engineering from BU and a master’s in computer engineering from WPI honed the intellectual rigor necessary to understand the underlying complexity of AI; an MBA from Columbia provided the framework that Alex uses to guide his team. His first job at Lockheed Martin worked with software for unmanned aerial vehicles and radio/wireless signal processing. This introduced him to the complex and hand-written algorithms that would later be consolidated and automated by deep-learning tools. Later on, Alex would join his first start-up as a founder at Visual Revenue, which built a real-time predictive analytics platform for editors to help them manage content and increase audience engagement. Visual Revenue was eventually acquired by Outbrain in 2013. With that exit, Alex took the next step on his current path of serial entrepreneurship.
Considering Alex’s background in the business of AI, I was fortunate to spend some time chatting with him about his experiences in the field, challenges he’s faced, and advice for budding data scientists like myself. Below are some of the questions we touched upon and the responses Alex provided. Please note that as I did not record our conversation, the responses are written from memory and notes, which were then sent to Alex for clarification and any addenda.
Build better voice apps. Get more articles & interviews from voice technology experts at voicetechpodcast.com
What did you foresee in the AI industry? What came true?
The democratization of data science — you don’t need to be a PhD or masters to be able to work in data science. Every quarter it becomes easier and easier. Big data was a specialized skill set, but now companies such as Amazon Services has given access to a whole suite of tools previously accessible only through manual programming. Services such as MongoDB and Hadoop provide the tools for distributed computing.
How do you pick projects to tackle?
I like to seek out an engineering challenge. I look for the impact and value we can unlock with the broad base technology. Work in a related field, and in a field where you can apply AI. A large repository of data — being able to apply analytics and algorithms to the data challenge. Fundamental data — how do you get access to large (high volume), well labeled data? Where can you add value?
Applying vision — image recognition — is the challenge I’m working on at AiCure. We are collecting data on patients and using image processing as a diagnostic tool. The platform will also provide interactive patient support and engagement, as well as patient data insights for clinical trials and doctors.
What are some of the challenges you’ve faced in developing the technologies/companies?
There were much more challenges in the beginning, mostly with the infrastructure and framework. No Tensorflow, PyTorch, AWS, etc., as these technologies and services was more nascent than today. Even 5 years ago, people didn’t have that much experience with it, as it was very niche knowledge. Now it’s less of a black box. However, finding the right talent continues to be a challenge considering the available pool of candidates and the number of companies looking to add qualified employees.
What’s the makeup of the data science team at your company?
We are more about cross functional teams: front end, back end, data science, and design. The idea of matrix teams — a team that is made up of different members from different functions. This allows for easier flow of information and streamlines the production process as all relevant stakeholders are at the table.
What do you look for in a candidate — specifically one for data science?
We are focused less and less on fundamental math background, and more and more on software development experience in building models. But they should have the fundamental data science understanding; how to tune, scale, and adapt to the problem we’re attempting to solve. We try to hire PhD’s and candidates with a master’s. However, the industry is opening up to qualified applicants who’ve gone through a coding boot camp, as it’s more important to be able deploy models.
Do you have any recommendations on job search and networking as a budding data scientist?
I see two manageable routes to get into data science. One is a less funded or resource strapped startup, where they may not be able to compete for candidates like Google or Amazon can. The other is a data rich company, which may be more willing to take a risk on a boot camp graduate. A large company, say in the retail or marketing sector, may have access to a lot of data but can’t compete against other tech companies in attracting candidates. Build your resume at a larger company and leverage the experience to move to other, more desirable companies. Go to meetups. Attending or competing in hackathons will help you stand out. Kaggle is a great resource for competitions with large datasets and a platform to distinguish yourself among your peers.
A big thank you to Alex Poon for allowing me a peek into the AI business world, and for the advice kindly given. I look forward to the challenges the industry will present going forward.
