Unleashing the Power of Data: How Applied Data Science is Revolutionizing Industries

data-science-in-production-building-scalable-model-pipelines

Applied Data Science
Applied data science and product data science will be covered in detail in this lesson.

We'll cover the following

Product data science
Applied data science
Need for applied scientists
Tools for applied scientists
Data science is a broad discipline with many different specializations. One distinction that is becoming common is product data science and applied data science.

Product data science
Product data scientists are typically embedded on a product team, such as a game studio, and they provide analysis and modeling that helps the team directly improve the product. For example, a product data scientist might find an issue with the first-time user experience in a game and make recommendations, such as which languages to focus on for localization to improve new user retention.

Applied data science
Applied data science is at the intersection of machine learning engineering and data science. Applied data scientists focus on building data products that product teams can integrate. For example, an applied scientist at a game publisher might build a recommendation service that different game teams can integrate into their products. Typically, this role is part of a central team that is responsible for owning a data product. A data product is a production system that provides predictive models, such as identifying which items a player is likely to buy.

Need for applied scientists
Applied scientist is a job title that is growing in usage across tech companies, including Amazon, Facebook, and Microsoft. The need for this type of role is growing because a single applied scientist can provide tremendous value to an organization. For example, instead of having product data scientists build bespoke propensity models for individual games, an applied scientist can build a scalable approach that provides a similar service across a portfolio of games.

At Zynga, one of the data products built by the team of applied scientists was a system called AutoModel, which provides several propensity models for all of their games, including the likelihood for a specific player to churn.

svg viewer

Tools for applied scientists
There have been a few developments in technology that have made applied science a reality.Tools for automated feature engineering, such as deep learning, and scalable computing environments, such as PySpark, have enabled companies to build large-scale data products with smaller team sizes.

The need to hire engineers for data ingestion and warehousing, data scientists for predictive modelling, and additional engineers for building a machine learning infrastructure are reduced. We can now use managed services in the cloud to enable applied scientists to take on more of the responsibilities previously designated to engineering teams.

svg viewer

One of the goals of this course is to help data scientists make the transition to applied science, by providing hands-on experience with different tools that can be used for scalable computing and standing up services for predictive models. We will work through different tools and cloud environments to build proof of concepts for data products that can translate to production environments.