Inside the Labs: the Power Within Network

June 06, 2017
By Chris Volinsky

Slide 0 Slide 1 Slide 2
Slide 0 Slide 1 Slide 2

Today, we have dozens of machine learning projects at AT&T Labs, addressing areas from proactive customer care to entertainment. In other words, AI and machine learning are woven into every day efforts. With it, we hope to one day operate more efficiently, better understand how customers use our products, and supercharge the customer experience.

While we’ve been working in AI for some time, the technology is still developing quickly. The way we use it today will inform the models we build tomorrow.

It’s important to have the right people working with the technology behind-the-scenes. And one thing I appreciate about my team is the range of perspective we have. There are multiple worldviews embedded in our work because of our range of expertise, age, gender and background.

To offer a better picture on where we are and where we’re going in our work with AI, I asked a few members from my team for their insight on the value AI adds to our business and what our future looks like in machine learning.

Why is AI integral to AT&T’s success?

Cheryl Flynn, Senior Inventive Scientist

PhD in Statistics

Years at Labs: 2.5

“We have access to so many types of data at AT&T. We can’t sit down as humans and look at all of it. We need tools like machine learning and statistics to help us gather insights from the data and think of innovative ways to solve problems.”

What roles does machine learning play in your job?

Wenling Hsu, Lead Inventive Scientist

PhD in Information Systems

Years at Labs: 20

“Cheryl and I worked with multiple teams across the business to successfully deploy our HelpCenter application. During the development phase, we pulled together the initial proof of concept and the engine was created in two days. The initial deployment was a success. This new functionality prompts visitors on ATT.com to explain their problem in a chat window, then uses predictive models, business rules, and Natural Language Processing (NLP) to route them efficiently to the appropriate chat agent group. Before now, visitors had to go through a self-guided navigation on various Entertainment Group chat launch pages. To build and deploy the HelpCenter solution, our team at AT&T Labs analyzed data from chat sessions and leveraged machine-learning methods to build initial models that determine customers’ intentions.”

Zhengyi Zhou, Senior Inventive Scientist

PhD in Applied Mathematics

Years at Labs: 1.5

“Machine learning is indispensable for my job. I use it in many stages of solving various business and societal problems, from design to implementation to evaluation. Recently, I’ve been involved in a number of anomaly detection projects. For example, for customer care, machine learning helps us monitor some various aspects of care performance and alerts us of preventative opportunities.

Another example is in using mobility data for finding anomalies in urban traffic. Urban planners can use the information to improve city transportation and infrastructure. We’re also looking into detecting, organizing, and understanding alerts in our network infrastructure and application layers for optimized operation.”

What’s on the horizon for your team in the year ahead?

Ann Skudlark, Director

Master of Business Administration

Years at Labs: 33

“I look forward to supporting AT&T Entertainment Group. We’re using a machine learning approach to better understand customers’ viewing habits, so we can recommend movies and TV shows they might not have considered before.”

DeDe Paul, Director

PhD in Mathematics

Years at Labs: 30

“We want to continue using human intelligence and machine learning to get the customer experience where we want it to be—effortless and proactive. We want to anticipate their needs. If we can make people’s lives even a fraction easier because of something we did, it’s a joy.

With all the data we have, we feel we should be able to do that now. Finding the right experts to explain the technical details is part of the challenge. Data alone doesn’t solve problems, people do.”

Filter by:

  • Most Popular
  • Author
  • Tags

Archives

Link to archives