On February 25, 2021, ADP unveiled an exciting and groundbreaking new payroll app called Roll™ by ADP® to help small businesses run payroll anywhere, anytime, quickly, and compliantly with no experience needed. The app’s artificial intelligence-backed conversational interface allows users to complete payroll on their mobile phones in less than a minute simply by texting, “Run my payroll.” Full press release. Watch a quick video of the app.
We had a chance to catch up with Roberto, a Machine Learning developer and one of the brilliant minds at ADP’s Brazil Innovation Lab in Porto Alegre.
Congratulations to you and your team on the launch of Roll™ by ADP®! We’d love to learn a bit about you. How long you’ve been at ADP, what brought you here, and what do you do here?
Yeah, sure. I’m working with machine learning here and part of Brazil’s Innovation Lab. I’ve been with ADP for three and a half years, so I started in 2017. I worked previously at HP, you know, the printer company, right? In their research lab.
I came here to start building the chatbot—a product complete within itself. A system where we can leverage the intelligence to make life easier for people that are using it. I’ve used chatbots, and sometimes they can be painful. Our job is to take the pain away. During development, we closely followed what our clients saw in production and what they said. When they are happy, that made us happy. We tried to understand our pre-production clients, make sense of what we learned, and iterate improvements before he launched.
So, our team is global and is split between here, the US, and India. We have about 13 people in Porto Alegre, but only four are working on just machine learning. We have around 32 people in Roseland, New Jersey, and about 20 colleagues in India. Our job here is to take care of the chatbot and help customers when they have questions. It’s kind of like using Alexa or Siri. When users ask questions, the AI is doing other things while trying to reply.
We’re also trying to extract insights from what customers are doing. For example, when you hire someone, we get the information behind the scenes, and then we do some tricky calculations to assist. The bot checks on things like gender and pay equity and offers data-driven insights to the client. For instance, in this location, you should offer a higher salary.
Tell us a bit about your career journey.
Sure. It’s a little bit messy. I’m an electrical engineer and worked a little bit in the automotive industry. I started as a hardware engineer working for Johnson Controls on a project for Fiat. Then I moved to a semiconductor company as an engineer and spent some time there. After that, I decided to move into technical marketing.
From there, I decided to get a master’s degree in Technology Management. I’ve lived with my wife in Lausanne – Switzerland, for two years. That was the initial plan. Then I got a job at Texas Instruments in technical sales. We stayed three more years before moving back to Brazil in 2015 and getting a job with HP. That was a big shift. I went from technical sales to software engineer. I had a colleague there that was working on machine learning. I fell in love with it, and I studied more about it. Then I got this opportunity at ADP to work 100% on machine learning. That’s why I came here. We pay 1 in 6 people in the US. There’s a lot of data here and good stuff we can do with it. So, I’ve been here since 2017.
I’m 41, almost 42, now. I have a daughter (Gabriela), she’s one year old. She is definitely my biggest project!
What still excites you about working here?
The team still energizes me. Before the pandemic, I enjoyed working with people globally and meeting the US teams in person when we still could travel to New York. We are trying to build this culture of applying available technologies and bridge the gap between open source and what folks in academia are doing with practical, real-life applications in our product, Roll™ by ADP®. Using this outside perspective, we filter what makes sense into our products to mature our technology.
I think the dynamics and openness of the machine learning domain are really driving the market right now. There’s a lot available in open source, and it’s our job to be up to date on the latest developments. It’s an exciting time to be working in machine learning.
Tell us a little about your project.
We beta tested with clients for almost two years. Last year, we did many internal demos based on our work with a gourmet ice cream company recommended by our Business Anthropologist, Martha Bird. We expanded and started working with our Small Business Services group and built our client list to 70 before we launched in February.
As we scaled for our GA release, we matured the product using input from a small number of clients. ADP’s executive team was happy with the product, and yeah, we hope people like what they see. As I mentioned, I go into production logs every week and see what customers are saying. Sometimes you get some nice comments, which is lovely, right? People talking to the chatbot and just saying, “Thank you!” I love seeing that. Martha measured pre-release net promoter scores (NPS), and they were really good. But we will keep the ball rolling and bring new features to future releases.
If someone asked you why they should choose ADP over other tech companies, what would you say?
I can say one of the things comparing ADP with the other companies where I’ve worked, and maybe it’s just specific to our product or my leader, but something I value a lot is openness. When I worked at other companies, there were a lot of layers. I think people are pretty open here also in terms of technology choices. I know that engineers like to experiment and test to see if stuff works. We try to do that here, experiment with things.
We are shifting from a service company to a more technology-oriented company. Here in Brazil, we are trying not only to apply technology but also to share ideas across the company. We created a machine learning discussion group. There are about 12 of us. We discuss papers, review articles, create challenges to learn new skills. We sometimes do presentations, attend or present at conferences. Everything is online today, which makes it easier. We get to exchange ideas and nurture our learnings across teams. We’ve discussed starting to produce some technical articles, and I’m happy that we can use the tech.adp.com blog to share them in 2021. I wish I had more time to write, but I don’t have as much time with my little one.
We also did our first internal developers conference in 2020, and I presented Uncertainty in Deep Learning. It was an amazing experience, great to share, but also to get feedback.
When I mention that we do these things during interviews and other things we are trying to do, candidates like this. I know in some companies people work in silos, but you cannot do that here in Brazil. We share as much as we can here. The openness I mentioned, it’s important.
Above, you mentioned exploring open-source and academia. Are there any projects outside of ADP that excite you right now?
Great question. Yes, there’s an open-source project called Open Mined and a course I’m interested in related to privacy with machine learning. The program is called “The Private AI Series.” Facebook is one of the course sponsors. They have a framework behind the scenes that helps people take care of customer privacy. In case you are interested, here’s the link: https://courses.openmined.org/.
Our team also continues to study and review new technologies. We are following Harvard CS224W online for graph neural networks and Causal Inference (lots of interesting applications will come out of this domain for sure!). For neuro-linguistic programming (NLP), we follow a vibrant startup and open source community called Huggingface. (https://huggingface.co/).
One last question! If you could advise your younger self or someone starting their career, what would you say?
Be inquisitive. Study. Help others.
Thanks for your time, Roberto!