Domino has helped Coatue increase productivity and achieve significant operational savings.
Data scientists run experiments asynchronously. “You can have multiple R Studio sessions or Jupyter notebooks open, so you test more things in parallel, reducing idle time,” said Izydorczyk.
Executing models is easy, self-service, and scalable. Users can quickly and self-sufficiently spin up environments that meet their needs and preferences; practitioners can seamlessly run their code on powerful AWS compute resources.
Employees are immediately productive; there’s no learning curve beyond existing data science skillsets needed to work in Domino.
Everything is reproducible, validated, and auditable. Because Domino saves every result generated by every string of code, results can be reproduced and validated multiple times, by multiple people. This is important for performance, as well as strong record-keeping and compliance.
Coatue’s data science team sought tools that would help them improve their quality and pace of work by:
Accelerating the research cycle. “With data science, you always have to try many things. It’s hard to just theorize on what the best answer should be,” explained Izydorczyk. “Most people write a short piece of analysis, run it, write another short piece, run it, write a third short piece, run it, and as a result, the speed of their iteration and development is bottlenecked by that run time. Data scientists need tools that let them run multiple experiments simultaneously.”
Deploying models into production. Before Domino, there was a lag between prototyping a model and putting it into production. “You’re in a perpetual state of having old code running in production. There’s a lot of value in being able to deploy immediately,” said Izydorczyk.
Coatue has deployed Domino in AWS as a central data science platform for both experimenting and deploying models. Domino tracks what models ran, when, and by whom. There is a system of record tracking the generation of new models throughout deployment. This provides clear statistical evidence of custody throughout the model management process.
Coatue has two main technical use cases for Domino:
- Developing models using the R programming language. Coatue uses Domino both for the ad hoc exploratory phase of model development, and for more rigorous testing.
- Scheduling analyses and models to run on a routine basis.
Beyond the aforementioned platform-based gains, Coatue appreciates the following benefits of working with Domino:
Customer support has been fast and thorough, investing effort to address unique feature requests and challenges.
Employee appeal: Domino allows data scientists to focus entirely on their substantive analysis, using the tools they’re comfortable with, and without having to worry about the underlying infrastructure.
1 Information presented herein regarding Coatue Management is for illustrative purposes only and does not constitute an endorsement by Coatue of, or a recommendation to invest in, Domino.