Python Workshop with TIAA

November 16, 2019 - 12:00 PM to 4:00 PM
UNC Charlotte Center City Room 1104
Productionizing with Flask

 

Instructors: Ahmane Curry-Muir, Kyra Koch, Stephen Rohrer, Garrick Stott

Date/Time: November 16, 2019 12-3:30pm (Lunch starts at 12:00)

Description

In this workshop, we aim to train participants on how to build the bread and toppings which are often neglected in the sandwich of data science. We will cover pulling data from an API, building/storing a model, creating an API for model consumption, and setting up model monitoring tools to prevent model drift and maintain faith in the model. The focus of this talk will not be on any kind of model development. More on the less interesting aspects of model development which are less frequently taught and less likely to be learned prior to entering the workforce.

Prerequisites

A basic understanding of Python and the Jupyter notebook environment is required. Familiarity with HTML will be helpful.

Workshop Participation

Students will be expected to participate by completing in-workshop activities, working in groups when possible, and asking questions when they have them.

Requirements for students
  • Web Browser
  • Python 3.3+
  • Jupyter Notebooks/Jupyter Lab
  • Downloaded copy of the provided content
  • The following packages installed on their machine
Python Packages Used
  • Scikit-learn
  • Pandas
  • Numpy
  • Flask
  • Statsmodels
  • Matplotlib
  • Plotly
  • Json
  • Seaborn
  • Lime
  • Requests
Outline

Topic

Topic Description

Host

Time Allotted

Getting Data from an API

Covering the basics of APIs and GET requests, common API status codes, JSON, rate-limiting, and an example modeling process using the Wikipedia API

Ahmane Curry-Muir

1 hour

Building a Flask API for Model Consumption

Storing and using stored models, HTML basics, Flask basics, creating an API endpoint

Kyra Koch

1 hour

Production Model Monitoring

Approaches to production monitoring, statistical methods for production monitoring, visualizations for production modeling

Garrick Stott

1 hour

Buffer time

30 Min

 
Goals
  • Describe the process from pulling data via an API through to model monitoring (skipping the meat in the middle of building a model)
  • Focus on the factors that enable a successful model deployment and maintenance
  • Understand the difference between an outlier and an anomaly
  • Synthesize their own production model monitoring system according to the project’s needs
Objectives
  • Pull data from an API data source
  • Build a Flask API for model consumption
  • Build an example model monitoring dashboard for future projects
Feed to: