Streamlit Data Explorations

This work for a client involved rapid prototyping of internal data exploration apps leveraging on a huge Snowflake database as resource and Streamlit as a tool to produce interactive visualizations.

One exploration involved diving deep into a dataset of socioeconomic, demographics, personal, and consumer information (Infutor). The Snowflake table had 250m rows, so random samples were extracted and processed in real time, from user interaction, with the help of polars, a dataframe library that can be many times faster than pandas.

Fast interactive exploration of the top consumer interests of any combination of demographics and geographic categories.

Another exploration involved a collection of tables from Cybersyn, comprising multiple timeseries from the Bureau of Labor Statistics, the Federal Reserver Economic Data, USPS addresses and address changes. Here, the solution involved interactively querying Snowflake to produce up-to-date plots with several parameters that the user could modify –i.e. normalized or unnormalized variables, seasonal adjustment, custom aggregation of industry types, and arbitrary combinations of cities, metropolitan areas, or states.

Timeseries of normalized employment for a selection of industries.

Stacked barplot of selected industries and their share of non-farm payrolls over the years.

Timeseries of net migration in a chosen city with the detail of inbound and outbound address changes.