When it comes to organizing your teams, how do you make sure you’re maximizing the capabilities of your data scientists within the enterprise? What is the best way to support innovation that leads to value?
When organizing data skills, there is a continuum from fully centralized to fully distributed, and I think people have very strong views on the right and wrong. My view is very pragmatic and that it needs to be tied to that organization. Some organizations are, by their culture, more distributed. Some are more centralized.
In some, you might need to have the data analysts and data scientists embedded in the business team structure with some kind of centralized support. Embedded analysts get closer to the data and customers and have greater control. More centralized teams, you have slightly higher efficiency because you've got less people to train and organize and there's less tools but the teams are less connected to the end users and data. So I think there's pros and cons of both. I don't think any are the right or wrong answer.
The larger and more sophisticated organizations are using more distributed data analytics teams. I think that comes with a level of maturity and the smaller or less sophisticated start with a centralized model. So I think it's always right to go from centralized to distributed, but I think it depends on the organization.
Content you might like
For data lakehouse platform utilization, our use cases vary between 1. respond to smart applications' data enquiries and 2. AI/ML data exploration. The first use case type mandates low latency responses, while the second consumes computational resources for long periods. Should we create two different lakehouse platforms to serve both use case types?
Context is more important than datum.
Strongly agree16%
Agree57%
Neutral21%
Disagree4%
Strongly disagree
View Results
As we prepare to launch Agentic products, we are exploring WebSocket implementations for streaming. We are primarily developing in AWS and both AWS AppSync and API Gateway require significant development effort. Are there any lower-code solutions or services that we could leverage to help speed up backend deployment and assist in managing the growing number of API's we are creating?
Who do you involve in defining fairness for models?
Data scientists11%
Business or product owners46%
Compliance or risk expert25%
Other data roles13%
Other non-data roles
I’d like to see the poll results4%
View Results
Is anyone using the company KeystoneAI for your data science or AI work? Wondering what your thoughts are on the company and if they can scale a solution for an enterprise.
What sets us apart?
No selling.
No recruiting.
No self promotion.
Read Our GuidelinesTrusted peer advice and insights for technology professionals.
When organizing data skills, there is a continuum from fully centralized to fully distributed, and I think people have very strong views on the right and wrong. My view is very pragmatic and that it needs to be tied to that organization. Some organizations are, by their culture, more distributed. Some are more centralized.
In some, you might need to have the data analysts and data scientists embedded in the business team structure with some kind of centralized support. Embedded analysts get closer to the data and customers and have greater control. More centralized teams, you have slightly higher efficiency because you've got less people to train and organize and there's less tools but the teams are less connected to the end users and data. So I think there's pros and cons of both. I don't think any are the right or wrong answer.
The larger and more sophisticated organizations are using more distributed data analytics teams. I think that comes with a level of maturity and the smaller or less sophisticated start with a centralized model. So I think it's always right to go from centralized to distributed, but I think it depends on the organization.