You can adjust all of your cookie settings by navigating the tabs in this window.
Before arriving at Nanigans over two years ago, Sharat Chikkerur spent five years at Google and Microsoft writing and deploying machine learning algorithms. Not a bad way to hone your skills.
Sharat says that experience — particularly his work at Google building large-scale machine learning systems that predict click-through rates for ads on the search engine giant — gave him a great foundation for his role as Nanigans principal data scientist.
We recently sat down with Sharat to discuss why data science careers are exploding and why the ad tech industry — and Nanigans — are tailor-made for data science.
I’m part of the optimization team. I started off here building machine learning models for the Facebook and Twitter side of the Nanigans product. These are models that predict click-through rates, action rates and other values for our enterprise-level advertisers such as Rovio, Wayfair and Rue La La.
I recently shifted from the social advertising side of Nanigans into the display side of the platform and the challenges there are different.
We build fewer models for display than Facebook, but each model handles several million events per day so the challenges are about achieving good accuracy while scaling. My main responsibility is to ensure the accuracy of these models and figure out which ones to build and how.
It’s been an interesting career path for many years, it was just called something else — maybe economist, statistician or machine learning engineer. But now that machine learning and AI [artificial intelligence] have become more mainstream, there are more data science opportunities.
Clients benefit directly from the accuracy of the models we build. So a data scientist at Nanigans is a high-visibility role at the center of the business.
Companies like Google and Facebook realized early on that data drives profit. Ecommerce and retailers are realizing that too, which is why they are turning to Nanigans to help them better understand and optimize their marketing spend.
In terms of what Nanigans does for its clients, we do broad optimization. So in addition to building machine learning algorithms, we need to know about control systems and economic theory too. It’s versatile role. There are always interesting problems for a data scientist to help solve that go beyond just writing or using machine learning algorithms.
Also, Nanigans has some of the largest advertisers in the world as clients. Almost a billion MAUs [monthly active users] are measured through our web/app tracking so we generate lots of data. In some cases, we have more insight into the clients’ users than the clients themselves. All of these clients have different volumes of data and modeling complexity, so a data scientist would have many diverse problems to address.
I think it’s mostly because in online advertising profit is tied directly to the machine learning models you build. Data scientists and optimization engineers have a direct impact on the company’s bottom line.
That’s very true at Nanigans. Clients benefit directly from the accuracy of the models we build. So a data scientist is a high-visibility role at the center of our business. At another company in another industry, data scientists may be there to give insights to upper management or do post-hoc analysis. But in ad tech and at Nanigans, data science drives the business.
As I mentioned, we help ensure ROI to advertisers, but to do that we must help address a few core issues first. Nanigans operates on multiple ad exchanges — think Facebook, Google, Twitter, Instagram and Rubicon, to name a few. A client has to figure out how to allocate their media budgets on these exchanges and then the next concern is how to efficiently bid on them. A client might bid higher for male vs female, young vs. old, week day vs. weekend.
These decisions are model-driven. Data scientists build a lot of models to establish what the return on investment is. At the same time they’re operating in an exchange in a competitive environment. So there are economic decisions to be made amid the pressure of competition.
We have almost daily interaction with the CEO and the rest of the executive team. And because we’re developing expanded functionality for display advertising that has a lot of data science needs, there’s tons of interaction with clients too.
Just to bid on Facebook, we build about tens of thousands of models per day across all our clients. On the display side, you have to listen to the bid and figure out how valuable it is within milliseconds. Again the volume of data is gigantic. The models have to handle tens of millions of data points per day across all our clients. We have to make choices about what kinds of models are suitable for that. That’s a big part of the data scientist’s job.
Well, first, I have to mention we’re offered unlimited vacation time! We also have free lunches from different restaurants each day of the week. Our office is in the heart of Boston near Downtown Crossing. Although we’re still a venture-backed tech company the work-life balance is better than it is at larger companies.
In terms of data science at Nanigans, everybody is in tune with what we’re doing. We have almost daily interaction with the CEO and the rest of the executive team. And because we’re developing expanded functionality for display advertising that has a lot of data science needs, there’s tons of interaction with clients too.
As far as the people go, there’s a nice combination of engineering, sales, ad ops and client-facing teams. It’s a healthy variety of personalities, which keeps the culture fun and diverse.
Nanigans’ data science team is looking to fill three data science roles in our Boston office to support our customer success and core tech engineering teams. Interested? Check out the links below.
Our latest report has exclusive data and essential insights from Q4 2017 that can make or break your advertising ROI on Facebook.
Download the Report