The machines are learning -- you should be too

Machine Learning has advanced far enough into the mainstream lexicon that if you were to ask a random cross-section of the population to provide their thoughts on the subject, you’d be bound to receive a broad array of opinions. But if you were to follow up and ask them what specifically machine learning encompasses, and how the science behind it works, you’d probably be met with plenty of shoulder shrugs and blank stares. I don’t know, I guess it’s where people program computers to think and then you hope they don’t turn on us, like in those Terminator movies?

In fairness, most people at least have the tacit understanding that the online ads, news stories, and suggested shows to watch are all based on some sort of behind-the-scenes machine learning programming. Or put another way, they’re keenly aware that their devices are trying to make sense of their past clicks and search histories to predict and influence their future behaviors. But generally speaking, most people have no idea how it all works, the scope of its utility, or where this field is really headed.

When it comes to machine learning, the public might as well be audience members at a Vegas magic show, cognizant there must be some sleight of hand at work but passively watching in amazement while waiting to see the next cool trick performed.

Now imagine yourself as a CTO. Instead of a random sampling from the general population, you are polling your entire organization. How confident are you that your people–everyone from the engineers to the product managers, agile delivery leads, recruiters. marketing managers, and all the rest–are genuinely machine learning fluent and proficient? You know machine learning isn’t an illusion and your organization’s future survival is predicated on your associates’ ability to be the ones up there on that proverbial machine learning stage, surprising and delighting the world with their prowess. Because if they can’t make the magic happen, it’s not much of a show, is it?

Here at Capital One, we’re the headlining act and we’re on 24/7/365. It didn’t take pulling a rabbit out of a hat to get here, either. Data is in our DNA and we were founded on the idea that we could do banking better by reimagining the power and potential of information. To that end, we’re fully  committed to helping our associates advance their machine learning skill sets and better prepare themselves for lengthy careers in tech.

As an Enterprise Learning & Development Contest Strategist, my role is to market and communicate Capital One’s Tech College program curriculum to our nearly 11,000 learners. Originally developed back in 2016 to help propel our historic cloud transformation, Tech College now offers thousands of training classes, providing technical skill development across all the essential tech disciplines, including mobile languages, cyber and dev ops practices, cloud engineering, and of course – machine learning and AI.

Upskilling 11,000 associates into the machine learning age might sound like a daunting task, but our robust and engaging collection of content combined with our inherently inquisitive associates makes for a fun and challenging endeavor. Through it all, I’ve found a particularly successful three-step approach for getting a diverse and sizable organization up to speed in a constantly evolving world.

Data, code, algorithms–machine learning can be an intimidating domain for engineers, let alone the laypeople in your organization. But here’s the deal: everyone needs to have a baseline understanding of machine learning and be fluent in its concepts in order to efficiently move a shared strategy forward. Machine learning should be every bit as ingrained in your corporate culture as having a working understanding of the cloud. Even if your role doesn’t require you to build and deploy apps, you at least need to understand the ecosystem to be on the same page and speak the same language as your engineering colleagues.

One of the crown jewels of Tech College’s machine learning catalog is the ML 101 video series. This offering was specifically designed to be an approachable entrypoint into the world of machine learning. Capital One’s Learning Program Managers leveraged their top data engineering colleagues to gently explain the basics of machine learning--including a general overview, the different types of machine learning, the machine learning life cycle, and the how and why of using machine learning responsibly. It’s almost like a Babbel or Rosetta Stone for understanding machine learning terminology and concepts.

In all, ML 101 is a 7-part series that clocks in at just under two hours which makes it ideal for binge watching. Obviously, you can’t teach a deep understanding of machine learning in 120 minutes, but that’s not the point. What you CAN do in two hours is give your audience a lay of the land to understand how and where machine learning fits in, and how they–in turn–fit into a machine learning world. And at the very least, they’ll be able to sound really smart and interesting at their next cocktail party. Speaking of supermodels, let me tell you about this incredible dataset I ran into the other day...

The next step of the process is to demystify machine learning as a concept and show practical use cases where it’s making a tangible difference in the business. Even after you’ve gotten your audience past the front door, you won’t gain any forward momentum until everyone begins to see machine learning as a real problem solving tool to be leveraged in their everyday roles. 

Tech College helped Capital One bridge this gap by creating and publishing internal case studies from innovative Capital One data teams who used machine learning to solve their relatable, but tricky, business problems. For example, think about how difficult it would be to read a downloaded transcript that didn’t contain any punctuation. Just a giant wall of text, where you’re never quite sure where one sentence or idea ends and the next begins. One of our amazing teams solved this issue by leveraging machine learning to incorporate a dataset that could train their production model to recognize when and where punctuation should be inserted into natural language conversations. Abracadabra, make this run-on sentence...disappear! Ta-da!

Once you’ve seen some concrete examples of machine learning’s utility, it makes it so much easier to begin imagining different ways to incorporate it into your toolbelt. You're holding the hammer, now start looking for some nails!

The final piece of the puzzle is providing your engineers with the confidence, skills, and knowledge needed to start getting under the hood with machine learning. At Capital One, Tech College created a customized learning path–a series of courses and hands-on exercises that teach you the fundamentals for becoming a machine learning practitioner. This is a next-level program that requires the learner to be fluent in Python and comfortable using GitHub.

Beginning with the building blocks of machine learning like statistics and linear algebra, to more advanced topics such as neural nets and deep learning, we provide our learners the tools they need to become genuine machine learning engineers. Many of these courses leverage Jupyter notebooks to present the content in a modular format with hands-on coding along the way. They cover theoretical and real-world concepts such as knowing the right time to implement machine learning to a problem or when a model is ready to go into production.

At Capital One, we know magical thinking isn’t a viable plan for achieving a world-class machine learning engineering platform. Instead, we work to ensure every associate has an approachable learning pathway that meets them where they are, and takes them to where technology is going. After all, it’s every bit as vital to invest in your talent to ensure they stay ahead of the accelerating pace of technology as it is to have the infrastructure capable of meeting tomorrow’s challenges.

It’s never too late for an organization to harness and embrace the power of machine learning. By helping your associates move past the intimidation factor, they'll begin to see machine learning for what it is: a better way to capture and make sense of the ubiquitous business data that’s all around us. Having a well-managed and unassailable machine learning organizational infrastructure that transforms those insights into better products and services becomes the authentic byproduct of a meticulously built culture of enabling associates to reach their full potential. You can’t get there with shortcuts or a magic wand, either. But when you’ve built a workforce like Capital One’s that’s equipped with cutting-edge skills, tools, and endless possibilities, it’ll just feel like it’s that simple to the outside world. And that’s the real magic.