Finding the perfect match with Machine Learning
Ateleris built a smart artificial intelligence for TestingTime to help find the perfect match from a large pool of candidates. The matching algorithm increased candidate matching performance by 250% by analyzing information from survey questions and predicting unknown properties.
TestingTime recruits people for usability tests, interviews, and other studies for many international clients. Each study has unique requirements, and it can be challenging to find people meeting these particular criteria.
Ateleris developed a machine learning system that finds suitable candidates from a pool of over 500’000 people, based on surveys that they filled out in the past. This artificial intelligence finds matches by analyzing correlations of related questions. For example, it automatically detects connections between income level and preferred car brands. With these connections, the matching algorithm can predict hobbies, preferences, and knowledge and select the candidates best suited for the given study.
The machine learning system was written in .NET/C# and ML.NET and ran as a self-contained Docker image in TestingTime’s Amazon AWS Cloud. This setup allowed easy integration into TestingTime’s existing IT infrastructure and processes. Additionally, it made regular smaller field tests and performance evaluations possible, which helped tune and improve the matching algorithm over time.
With the machine learning system’s help, the chances for a successful match increased by 250%. Moreover, the smart artificial intelligence algorithm continuously learns from new data and becomes more intelligent day by day.
Key Technologies/Terms
- Machine Learning / Artificial Intelligence
- .NET/C#
- ML.NET
- MongoDB
- Docker on AWS Elastic Beanstalk