As machine learning becomes a top of mind business strategy for organizations, developers need to understand their role in the overall workflow Credit: Thinkstock Developers are independent thinkers, often preferring to work on their own projects and shying away from cross-team collaboration. Yet, as the pressure for organizations to implement successful machine learning strategies grows, the role of the collaborative developer is more critical than before. But why? Machine learning has come a long way since it was a novel idea in 1959 to automate processes. Today, the stakes are much higher for enterprises applying machine learning technology, so here’s an approach for developers to become more comfortable with their new role in the machine learning workflow. Establish common ground with data science teams Successful machine learning strategies require complete buy-in from all parts of the organization. Teams must be prepared to act on the data machine learning yields, and they must be ready to make decisions based upon that data. There’s a tendency with any kind of transformative tech to think of it as magic, and therefore assume that it doesn’t need support to help it function. But machine learning needs a team ready to back it up, and that includes a united effort from both data scientists and developers. Developers should first establish common ground and vocabulary with data scientists. The two don’t speak the same language, and they don’t often have the same goals and objectives, but understanding and agreeing on common tooling is key so models created by data scientists can be used and deployed by developers. Accept the changing role of the developer Instead of creating and deploying all the code, developers can now take the machine learning model and turn it into a production system. This isn’t the same traditional developer gig, but it’s just as important as before. Learning machines require an automated flow that goes from data to predictions, and reverse, collecting feedback from the effect of predictions in order to further train the machine. E-commerce sites are an example of this application. When you put something in your basket, the site recommends additional products. You need a great data science team and good algorithms to create a model for how people buy products. The developer enters the picture at the moment you have this model. Developers are required to turn the model into a component of the e-commerce site recommendation system; they make the move from data science lab to predictions possible. Turn business problems into machine learning algorithms Issues abound in the enterprise that can be solved by machine learning. To solve any business problem, you must first start with a business question. Some vendors speak too much about “magic.” In their eyes, you combine data, machine learning and cognitive artificial intelligence, and you get magic. That may be true someday in the future, but today, projects should start with a well-defined question, such as: Can I recommend products to my visitors? Can I predict who is going to leave me for competition among my current customer base? Can I predict next month? When developers are brought into the fold to deploy models and improve applications that answer real business problems, enterprises see results from machine learning. Take advantage of machine learning behind the firewall The ability to create, train, and deploy applications that can identify incremental opportunities makes for a successful machine learning strategy. And nowadays, developers have more access to the tools needed to contribute to better in-production solutions so enterprises can make more accurate predictions, faster. Moving machine learning to a hybrid cloud environment improves lifecycle management and better serves all parties involved in a strategic machine learning strategy, supporting the critical role developers play in the machine learning workflow. Software DevelopmentArtificial IntelligenceEngineerData Science