Presented by:


Jim Nasby

Amazon Web Services

Jim discovered a passion for data and databases early in life, and was working as a database consultant before high school. Since then he has worked with everything from Foxpro to Oracle. For the past 15 years he has specialized in PostgreSQL and is an active member of it's development community. He specializes in tackling complex data challenges where data quality is critical.

No video of the event yet, sorry!

In data science there is often a huge disconnect between data scientists trying to create models and the application development team responsible for implementing them. Often this is made worse by data scientists using a different language for modeling than the application is written in. Modeling in Postgres eliminates this issue by allowing the model to be written in any supported language (such as python or R). In this talk I will present a modeling framework created for a client to meet these needs. - Data scientists can perform data discover and model creation in their choice of language - Gathering data is done in separate functions, so it's easy to use data-focused languages like plpgsql - The framework makes it easy to retroactively run the production model code against previous results - Once the model is validated it is stored immutably in the database, providing for robust regulatory auditing

30 min
PGConf US 2016 [PgConf.US]