Interview with Jason Shu, SVP, Data Science, Aki Technologies

Tell us about your role at Aki?

At Aki Technologies, we connect brands to audiences during their most receptive and relevant moments. Naturally, a lot of what we do relies on data and the smart interpretation of data, which is where my team comes in. We’re focused on optimizing the delivery of marketing to mobile users and innovating new methods for reporting on how well those ads affected desired outcomes.

We are proud of the Machine Learning algorithms we’ve built to understand users’ mobile mindsets, but AI is just part of what we do. We’ve built an end-to-end solution that enables us to build marketing strategies for clients and to look beyond the click to objectively measure how well those strategies work.

How would you define a ‘mobile moment’?

A mobile moment is a combination of circumstances that help marketers understand the mindset of a mobile consumer. It is no longer enough to simply look at the time of day and location to infer what a user is doing – the nuances of the always-on mobile experience require a deeper understanding. We’ve built models to help marketers understand how to deliver an ad that is relevant to the way that consumer is feeling during that moment.

How does Aki’s platform help drive the ROI of marketing campaigns? 

By interpreting the mindset and receptivity of a mobile consumer during a given moment, Aki is able to predict when and how a consumer will respond to a marketing message. For example, if a consumer is in a “lean-back,” relaxed mode at home, Aki can optimize performance by delivering a rich media or video ad. Conversely, if the consumer is on the go, Aki can serve more cost-effective “reminder” banner ads. Aki uses a combination of deep audience insights and AI-powered targeting to ensure the best possible experience for the customer and the most efficient path to strong performance.

Which industry verticals does your product find the most resonance in and why?

Today’s marketers recognize that mobile advertising requires a deeper understanding of consumer behavior. Smart marketers are mapping out the moments that are most relevant to their brands and specific campaign objectives to drive greater impact and eliminate waste on poorly-timed impressions. Aki’s moment marketing science provides a framework for this, which is why it is being used with great success across different verticals, from auto to travel to finance to entertainment, restaurants, CPG. The list goes on.

CPG is a great example of a vertical that’s seen a lot of success, partly because moments are such a natural extension of “need states.” Automotive too, because the long consideration cycle requires smarter and timelier engagement. Pair this with Aki’s ability to measure foot traffic in real time, and it’s not surprising we’re seeing a lot of demand in both areas, among others.

In terms of martech, what direction do you see AI/ML moving in?

Machine-learning models love nice clean outcomes (or “labels” in ML parlance) which is why click-through rates have remained the darling of martech. For too long, the industry has blindly focused on how to optimize CTR without any thought about actual marketing strategy or what an ad was intended to do.

Clicks are a relic of the old e-commerce model in which getting someone to visit a website was the only way to make a sale. This doesn’t work in mobile. The point of mobile is to inform, engage, and remind—not immediately get someone to checkout. Brands are starting to understand this difference and are adapting their expectations. Martech has to respond to these new expectations.

I believe that we are going to see far more emphasis on developing models that identify the best ads to match the mindset of a consumer, without a care for whether they click on an ad or not. Obviously, this is much harder than optimizing on clicks, but it is at the heart of what Aki is doing with mobile moments.

What startups are you watching/keen on right now?

Deep Learning is all the rage in the AI/ML space, but its complexity has made it difficult for non-experts to easily prototype new ideas. Libraries such as Keras and TensorFlow have been the go-to tools to help simplify the model-building process, but even these tools have their limits. PyTorch is a new library developed by fast.ai, an organization teaching Deep Learning through the University of San Francisco’s Data Institute. This library expands the breadth of problems that can be solved, while also further simplifying the development and model training processes. I’m excited to see what marketing experts can do, given more time to solve problems and less time to worry about the nuts and bolts of Deep Learning models.

What apps/software/tools can’t you live without?

Search is a Data Scientist’s best friend. It’s the old adage: “If you have a question, there’s a good chance someone else in the room is wondering the same thing.” Why re-invent the wheel if someone else has already solved your problem? This is why I let candidates use Google when they are going through their on-site interviews.

Other than that, iPython is my most indispensable tool for doing analysis or prototyping. When it’s combined with a good Integrated Development Environment (I use both Spyder and PyCharm), I can crank out analyses really really fast.

Which technologies within the AI/ML ambit, do you believe have the most potential?

Advancements in Deep Learning are coming at an incredible pace. The unique things they are capable will result in smarter models capable of predicting outcomes far more impactful than simple clicks.

One word that best describes how you work.

With integrity.

What’s your smartest work related shortcut or productivity hack?

The internal framework that I built has significantly decreased the time it takes us to prototype ideas. It also drastically decreased the learning curve that new team members have to go through before delivering meaningful work. Our framework is simple yet powerful, and our data scientists love contributing to the code base as a way of easily sharing their own learnings and productivity hacks.

What are you currently reading? (What do you read, and how do you consume information?)

I’m currently reading “Thinking Fast and Slow” by Daniel Kahneman and it covers a lot of ideas that can really help data scientists improve the way they approach data and problem-solving. There’s a common assumption that good data scientists just need to know algorithms and be really good at math, but to generate real value you need to do more than simply pass data through models. A great data scientist needs to be perpetually curious, and then be able to deliver on that curiosity with exceptional critical thinking skills—Kahneman’s book provides a lot of insight to that end.

Besides books, my major source of information is actually my Linkedin feed. I’ve connected with so many thought leaders in the Data Science community, and I something new from them almost every day.

What’s the best advice you’ve ever received?

“Hire people with fire in their belly.”

Early in my career, I asked my boss why he chose me over some of the other (clearly more qualified) candidates he had in his pipeline. His response: “You had a raging fire in your belly. I could teach you skills you don’t yet have, but I couldn’t teach the other candidates how to be hungry.”

And while this isn’t advice as much as inspiration, at Aki, we often talk about the importance of grit. We’re working to address the limitations and frustrations in mobile advertising for both brands and consumers. It’s the kind of work that requires a deep commitment and a lot of resolve, and we’ve got plenty of both at Aki.

Tag the one person in the AI/ML sector, whose answers to these questions you would love to read:

I think everybody could learn a thing or two from Henry Humadi – Data Science Lead at Vungle. Henry and I served on the Advisory Board for the Data Institute at the University of San Francisco. He’s a tremendous talent and leader.

Thank you Jason! That was fun and hope to see you back on MarTech Series soon.