Developers and modelers face challenges when finding and validating data, collaborating across groups, and transferring work to an enterprise platform.
Using a self-service, on-demand compute environment for data analysis and machine learning models increases productivity and performance while minimizing IT support and cost.
In this Q&A, Joe Madden, Senior Product Manager at SAS, explains why a workbench is an ideal environment for developers and modelers.
Q: Why use a workbench?
Joe Madden: A developer workbench is a robust, cloud-native environment that streamlines many types of analytical development. It provides a flexible environment to experiment, iterate and deploy models efficiently and cost-effectively, while balancing data security and interoperability that meets diverse user needs.
Q: How do developers and modelers benefit?
Madden: Developers and modelers benefit in several ways, including efficiency, scalability, collaboration and quality of results. It simplifies the process of writing, testing and optimizing code for machine learning models. Users can collaborate seamlessly to share code, insights and best practices. The environment handles large data sets and complex computations with quicker startup times, making it ideal for scalable projects.
A workbench also provides tools for building, training and evaluating models. You can experiment with different algorithms and hyperparameters. Users can track model versions, ensuring reproducibility and transparency.
Q: What about access to data and data security?
Madden: Access to quality data is critical and must be managed within the environment. Developers and modelers will not waste time wrangling data because data management is centric within the tool. Sensitive data stays within a firewall, which is important in highly regulated industries.
Q: How important is a multilanguage environment?
Madden: Flexibility is important. You hear about disparate data sources and programming languages. Developers and modelers might have a language of choice or a language they are most familiar with. A developer workbench supporting a multilanguage environment makes it easy. For instance, you can write Python code using very simple syntax. You won’t have to know SAS® if you learned Python in school. It benefits novice and experienced users. And users with different language preferences can work together using what they know best.
Q: Can this tool work with generative AI?
Madden: Sure. For example, a user can use a natural language prompt to generate SAS code that is ready to go with a click of a button. It will jump-start your code creation, but it is not intended to be complete code and will require some editing. This accelerates a developer’s productivity, but a human needs to be involved at every step to fine-tune your model to ensure meaningful, trusted results.
Q: Can you share how this could be applied in a real-world setting?
Madden: Imagine a retail company analyzing customer behavior. Developers and modelers can preprocess data, create features and build pipelines using a workbench. Modelers can then experiment with various algorithms, like XGBoost or neural networks to predict customer preferences. Users can drastically accelerate time to value to solve real-world problems, like improving customer satisfaction.
Q: What about applications across industries?
Madden: A workbench for developers and modelers has broad use and application. Think about a PhD researcher in higher education who is dealing with massive amounts of data and is on a tight budget. Time is money, and there is no time to wait for results to run over the weekend. Enabling more GPUs for a certain project or parts of that project when needed is huge. However, other parts of their workload might need medium-sized computing power, especially if they are doing exploratory work on a data set. Users can control compute resources and scale up or down based on project processing needs. This results in greater efficiency and cost control.
Q: What about the next generation of users?
Madden: A workbench is about customization and flexibility, which is attractive to the next generation. That means open source integration with multilanguage architecture to support disparate user needs. It allows users to self-provision an environment without additional IT support and choose their GPU/CPU server compute power to match the scope of the project. Ultimately, it gives the user flexibility, control and choice.
Faster, high-performance models deployed
A workbench tailored for developers and modelers accelerates model development and deployment in an on-demand, scalable, secure environment with trusted outputs.
