Sean Lensborn, CIO at DBRS
Financial services firms have been running their businesses on predictive models for decades. Bringing transparency into the modeling process and facilitating rapid development and collaboration has become paramount.
Efficiency: With data science teams geographically distributed across the world, some duplication of effort is inevitable. Overlapping efforts need to be identified and reproducibility should be made possible.
Governance: Financial institutions must delicately balance managing demands from regulators and stakeholders. Increased governance, need for transparency, and strict auditability requirements around models and their methodology have prompted institutions to focus investments in such areas.
Self-service: Data scientists have varied needs for hardware, infrastructure, and tools to be productive. Reliance on IT and Engineering to accommodate these needs within regulated organizations adds immense pressure and delays for both teams.
Model management and governance: Multiple ways to deploy, produce, or operationalize finished models, instead of driving data scientists to set up their own IT systems or go through arduous processes with application engineers. This includes deploying models to power scheduled jobs, reports, APIs, or dashboards in one place. Domino also provides a consistent baseline of non-functional requirements such as security, high availability, and so on, and as a system of record for all data science activity, offers transparency into assets and utilization across the enterprise.
Self-service infrastructure: Data scientists can do exploratory data analysis and model development without configuring and using their own compute resources. Data scientists can spin up high-powered workspaces with a single click, without needing help from Engineering.
Domino encompasses compute resources—as well as the languages, packages, and tools necessary for modern data science work—with controls and reporting around resource usage to administer or attribute costs. It will also automatically spin down an inactive environment to save utilization costs.
Reproducibility: Distributed teams of data scientists can discover past work and areas of expertise within the organization. Things that would have taken days to figure out on a local machine are resolved in hours by referencing something a colleague has already done, and employees learn from each other at a faster pace.
Demonstrate governance and relieve researchers from having to compile documentation. Domino makes it easy to show all the data and code that went into analysis, as it preserves full experimental records automatically—who was involved, what methodologies were attempted and discarded, when was it promoted to production, and so on.
Collaboration: Data scientists get hands-on involvement from across the team without anyone needing to replicate environments. Domino is language agnostic, so data scientists can work with their tools of choice and business users have the ability to access powerful data science insights using tools they’re comfortable with without infrastructure headaches.
Domino lets data science and quantitative research teams perform ad hoc experimentation and analysis, schedule runs to collect, aggregate, and cleanse data, develop financial models, deploy and maintain lineage of credit ratings, detect fraud, meet auditing requirements, and more.