MDS Employer Spotlight: Dialpad
Capstone Partner Profile
Within Dialpad, a provider of a cloud-native business communications platform, you will find that the majority of its projects are data science-oriented.
“We have a fast-growing user base who are continuously feeding data into our data warehouse. We use data science for various analyses of large-scale data that we would have to do manually otherwise,” said Matthias Lee, NLP Data Scientist at Dialpad and MDS Vancouver alumna, class of 2019.
Some of the data that Dialpad deals with include data from its Voice Intelligence (Vi™) team that works on Automatic Speech Recognition (ASR), Natural Language Processing (NLP), and Computer Vision (CV) to help its customers derive more from their calls. Within the Vi team, some of the data they are dealing with include audio data (.wav files), text data (mainly transcripts from business conversations), video recordings (.mp4 files) and user behavior data.
“We do a range of data science projects on different scales ranging from finding problematic training data or finding trends in data to developing ASR and NLP models working in the production environment,” said Chao Wang, Tech Lead, Data Analyst at Dialpad and another MDS Vancouver alumna, class of 2019.
As a team with a lot of ongoing projects and responsibilities, Dialpad is constantly coming up with many interesting ideas but they often give way to projects that have higher priority. Wanting a way to test some of these ideas without the pressure of a product deadline led Dialpad to collaborate with the MDS program and become a capstone partner.
“The UBC MDS program is one of the best data science programs in the world. Naturally, working collaboratively with MDS students would not only give us the opportunity to work on those projects that can bring long-term benefit to Dialpad, but also gives MDS students the opportunity to work with real-world big data and cloud computing tools and resources,” said Nathan Zhang, Manager, Data Analysis at Dialpad.
This year, Dialpad had three capstone projects with the MDS program. Two were for the MDS Vancouver program and the other was for the MDS Computational Linguistics program.
The first project had the MDS students detecting trends and anomalies in Call Center conversations from over a year’s worth of data.
“We are going to use the pipeline they developed to analyze the data on a larger scale and inform the feature development,” said Wang.
The second project focused on extracting acoustic features from audio and understanding which types of acoustic features are useful in predicting customer satisfaction. The new insights and results that the students provided from the project is going to be used to help Dialpad accelerate its development in its proof of concept.
“The students developed a possible approach to one of the central problems in a feature we are currently building, and also provided data analysis insights. We are going to use the pipeline they developed to analyze the data on a larger scale and inform the feature development,” noted Lee.
For the MDS-CL program, its project had the students identify the problematic human transcription used in ASR model training with little labeled data. Dialpad plans to integrate part of the pipeline students have built into its internal platform.
Zhang added that they wanted to work with MDS students on these capstone projects because of the fresh perspectives they can give to the Dialpad team.
“We were interested to know how people outside of our industry would approach our data,” said Zhang. “For the same reason - having been students ourselves - we know the pain of not having enough industry experience when looking for the first job. We wanted to provide this opportunity to decrease the experience gap for students at one of the best data science programs.”
They added that it was a lot of fun working with the MDS students, in particular, “it was great to see them developing creative solutions to different challenges they faced during the project, engineering novel features to improve their model’s performance and providing new insights.”
Lee, Wang, and Zhang encourage other organizations with interesting ideas and associated data to submit a capstone project proposal especially if they lack time or resources to work on those projects. They added that Dialpad plans on submitting a capstone project for the next cohort of MDS students.