7 ways SAS empowers startups with artificial intelligence and machine learning

0

SAS for startups

The startup ecosystem is dynamic and the flow of venture capital into tech is at an all-time high. Billions of dollars are invested in tech startups every year. Many tech startups market themselves as ‘powered by AI’ and pitch investors with buzzword laden phrases such as, ‘we leverage state of the art machine learning algorithms to do X’. There’s a significant push to incorporate AI and machine learning into their B2B and B2C software offerings. Are you one of these startups?

Rather than try to develop AI and machine learning strategies on your own, startups should consider SAS artificial intelligence and machine learning solutions which are developer friendly and ready to deploy. With SAS Viya, there’s ample opportunity to offer machine learning (ML) capabilities that will seamlessly integrate with the rest of the software systems that vibrant startups are building. Here are seven ways how using SAS Viya as a ML development platform could benefit your company:

1. Shorten time to value and increase speed to value -- significantly

Building comprehensive data preparation pipelines, robust ML models tailored for specific tasks, and model management are all quite involved and extremely difficult tasks. If you were to use popular open source tools alone, it would take a significant amount of time to mix and match the various libraries, computing infrastructure and technical staff to produce an integrated system that fits the unique needs of your startup.

SAS Viya is a powerful and comprehensive platform that integrates data preparation, ML model development and model governance (i.e. deployment, monitoring and maintenance). These robust capabilities allow a data analyst, a data scientist, or a developer to build and deploy production-ready ML models in a fraction of the time it usually takes to build infrastructure solely with open source software. This is a significant advantage for a team that wants to take product to market quicker, with powerful AI and ML capabilities.

2. Enable your data scientists and developers with a language agnostic AI/ML platform

Some developers understand the application requirements, but not the specifics of implementing open source ML algorithms at a production-ready level. Some data scientists understand the statistics and the domain, but might be unfamiliar with a certain ML development stack or a class of algorithms. For example, an R programmer may not be utilizing the same tool set as a data scientist who grew up with Python. At a time when data scientists and ML developers are in short supply, your hiring decision may be hampered by the choice of a programming language.

SAS Viya allows all technical team members to contribute to the data science and ML initiatives by functioning as a language agnostic ML development environment. The drag-and-drop interface helps to prepare data and create ML pipelines. Python or R users can use powerful ML algorithms implemented within SAS Viya inside scripts written in the language of their choice, and within the visual model building interface.

Model governance is a critical function in ensuring efficacy of deployed models, and SAS Viya facilitates the management of multiple ML models, regardless of the development language. Also, the entire platform is powered by Cloud Analytic Services (CAS), a computational engine that employs multiprocessing and multithreading to ensuress the maximum utilization of available computing resources. In short, SAS Viya unifies a talent pool with diverse data science skills, while facilitating rapid AI/ML development.

3. Utilize open source in tandem with SAS technology for a flexible development experience

As mentioned, SAS Viya allows Python and R users to employ algorithms implemented in SAS within a Python script or an R script. This allows a creative data scientist or developer to use SAS algorithm implementations (accessed through API calls) alongside functions from open source libraries or proprietary algorithms. Furthermore, you could include ‘nodes’ which encapsulate code written in R, Python, Lua or Java in a ML pipeline created with SAS Viya’s visual drag-and-drop interface.

Regardless of the web development framework used, ML models developed using SAS Viya can be accessed through API calls. Thus, embedding ML capabilities in a web application created with Flask, Django, Ruby on Rails, Laravel, or any other web development framework, is much easier with SAS. This embedding gives any technical contributor maximum freedom to develop ML capabilities efficiently and seamlessly, and integrate with a customer-facing web application.

4. Minimize development productivity loss in a world with high data scientist turnover.

Maintaining a core tech team in your organization is important, but difficult when data scientists and ML developers are in such high demand.

Assume the time to market for an early stage startup’s first product is one to two years (which is entirely reasonable, if not optimistic in some cases). The average tenure of a startup employee is 10.8 months. That means that, on average, the data scientist position opened at the launch of a startup is held by at least two individuals by the time the first product is launched.

On average, the cost of salary, benefits and training for each data scientist is over $139,840, depending on the geographical location. This is inevitable, and just part of the current industry landscape and options that exists for tech workers. But the additional productivity loss caused by the handoff to a new coder using purely open source frameworks can be significant. And while version control works well to some extent, the changeover always hurts.

With a structured and integrated ML and model governance platform such as SAS Viya, the flow of ML development is ensured and strengthens the team of core developers. Documented data preparation, ML pipeline templates (standard or built from scratch, based on each developer’s unique needs), and automated hyper parameter optimization features enable transparent and automated ML development.

5. Scale and adapt to a variety of application architectures

The runtime environment CAS that powers SAS Viya enables massive parallel processing and multithreading to efficiently leverage the computational infrastructure available to a developer or a web application in production. CAS can interface with Hadoop and function as a robust computational layer. SAS Viya can be deployed on premise or in any cloud service provider.

SAS Viya components can be containerized (e.g. dockerized), allowing instances to be spun up or down at the will of the developer, or according to demand, and can be orchestrated with container frameworks such as Kubernetes or Docker swarm. This provides data scientists the ability to explore SAS Viya’s data science capabilities in a sandbox environment and, more importantly, allows SAS Viya to be part of data intensive applications that follow microservices or serverless architectures.

Thus, SAS Viya’s ML capabilities can scale with data volume as well as design requirements, which is critical for integration with other software components. This further facilitates the embedding of ML-powered analytics in a web application, either browser based or mobile.

6. Minimize dependency issues

Open source software, by its very nature, exists due to community contributions. Some open source software is incredibly well supported, and that makes them ideal candidates for startups. Pairing a platform such as SAS Viya with such open source libraries would enhance the productivity of the data scientist and the quality of the product. However, some open source libraries, which may have been category leaders at one time, gradually lose support, sometimes surprisingly quickly. And when support wanes, the dependencies of such libraries may be too old to coexist with newer dependencies required to run newer open source components and the core application itself.

This issue can be circumvented by architectural changes, wrapping components in Restful APIs and perhaps containerizing, but can be very difficult to maintain. However, modifying and swapping out chunks of code to stay up to date when a poorly supported library is replaced with a newer open source tool can be even more difficult, particularly down the road.

Utilizing the ML and model management capabilities of SAS Viya helps the startup data scientist minimize, if not avoid these issues, as ML algorithms and critical features such as autotuning of hyperparameters are continuously updated by dedicated R&D staff at SAS.

7. Avoid issues in using certain open source tools in closed source software

Open source software is free to use, and you can see under the hood and play with the code as you like. However, not all open source software is the same. Each open source library or framework could be released under a number of different licenses. At times, depending on the license, software that utilizes open source software is required to be open source as well. This could lead to a number of issues for early stage technology startups, whose biggest asset and key differentiator is the code that powers the product. Thus, if a developer relies on open source software without carefully reading the license of each open source library used, and the startup uses the software in a manner not permitted by the license, there's a very real possibility of breach of contract and copyright infringement lawsuits.

What’s your advice for a technology startup who wants to add AI and ML to their product?  I welcome your insight.

If you’re interested in learning more about how SAS Viya meets the needs of the startup community, check out the Berg Data Solutions customer story, or visit the SAS Viya page.

Share

About Author

Avinash Sooriyarachchi

Solutions Architect - Artificial Intelligence and Machine Learning

Avinash Sooriyarachchi is a Solutions Architect with an emphasis on artificial Intelligence, machine learning and the integration of open-source technologies with SAS. Before joining SAS, he tried his hand at tech entrepreneurship by establishing a startup with a focus on Natural Language Processing (NLP), particularly for natural language querying. Currently at SAS, he focuses mostly on supporting opportunities related to applications of artificial intelligence and machine learning in mid-market healthcare, life sciences and technology organizations.

Leave A Reply

Back to Top