In this tutorial, we streamline model predictions with an API endpoint that accepts a raw string for inferencing. We also experiment with model signatures for a more customizable payload.
Here’s the GitHub repo with all of the code.
Around one year ago, I wrote Deploy a Keras Model for Text Classification using TensorFlow Serving for deploying text classifiers in TensorFlow 1.X. Since then, I’ve spent a lot of time migrating older projects to TensorFlow 2.X. While I found 2.X much easier to work with, I had a hard time finding the documentation I needed, especially around deployment and migrating deprecated V1 features for TensorFlow Serving.
I’ve gathered my learnings and will be demonstrating model deployment with a toy example.
If this is your first time working with TensorFlow 2.X, …
Tips to stand out in a competitive job market
This article is cross-posted on my website https://www.dscrashcourse.com/ alongside other guides we have been curating for aspiring data scientists.
The data scientist role is a popular career choice for anyone who likes to work with numbers and analytics. Once referred to as “the sexiest job in the 21st century” by Harvard Business review, the popularity of this industry has caused it to become oversaturated with job seekers and bootcamps.
With so much interest and competition, it has become harder for aspiring data scientists to stand out and get noticed in the job market.
So, what can an aspiring data scientist do to stand out?
First, it’s important to understand the reality…
Deploy an AI assistant using Rasa — from conception to Facebook — within an hour.
Rasa is an open-source conversational AI framework that uses machine learning to build chatbots and AI assistants. Today, I’m going to show you how to build your own simple chatbot using Rasa and deploying it as a bot to Facebook messenger — all within an hour. All you need is some simple Python programming and a working internet connection.
The complete code can be found here: GitHub Repo
The code was developed and tested in Python 3.7. Rasa currently only supports Python up to 3.8 (see here for updates).
The bot we build today will be very simple and will…
Set up GPT-2 demo in under 5 minutes using Streamlit
GitHub Repo: ml-streamlit-demo
Bringing a Machine Learning model outside of a notebook environment and turning it into a beautiful data product used to be a lot of work. Luckily, there’s a lot of tooling being developed in this area to make prototyping easier. A while ago, I came across Streamlit, an open source Python library for building custom web apps.
It’s quick and easy to get started and took me less than 30 minutes to build an app with a pre-trained model. Since then, I have been using Streamlit for prototyping models and demonstrating their capabilities. …
Share your app with the internet in under 5 minutes
If you are not working with an existing app, you can refer to my Iris Classifier FastAPI App for reference. I wrote an article about how to set that up (the Docker component is optional).
Note: I put the Iris app together really quickly to demonstrate how to set up FastAPI — it certainly does not adhere to best practices when it comes to serving model predictions
This is my GitHub repo for the Iris Classifier app. We will be using this as an example in this tutorial. …
Step-by-step guide for using Airflow + Docker to deliver weather forecasts to Slack
Tech Stack: Python 3.7, Airflow (1.10.10), Docker
GitHub link: All of the code can be found here.
Airflow + Slack
Slack is an increasingly popular chat app used in the workplace. Apache Airflow is an open source platform for orchestrating workflows. One of the biggest advantages to using Airflow is the versatility around its hooks and operators. Hooks are interfaces to external platforms, databases and also serve as the basic building blocks of Operators.
The Slack Webhook Operator can be used to integrate Airflow with Slack. …
Can our model accurately detect churn to help retain these customers?
Customer churn, also known as attrition, occurs when a customer stops doing business with a company. Understanding and detecting churn is the first step to retaining these customers and improving the company’s offerings.
Telco Dataset
We will be training our churn model over the Telco-Customer-Churn Dataset to predict the likelihood of customers leaving the fictional telecommunications company, Telco. This synthetic dataset was put together by IBM and includes a label indicating whether or not the customer left within the last month.
Goal: predict whether a customer will churn based on their demographic and service information.
Data Exploration
The exploration and modelling will be conducted…
Data Science Crash Course
The Ultimate Python library for working with Relational Data
Earlier this month, Edward Qian and I started working on a set of comprehensive lessons for aspiring Data Scientists, which can be found on our website www.dscrashcourse.com
I will be cross-posting slightly modified lessons to Medium to make them available to a broader audience. If you find these articles helpful, check out the site for more lessons and practice problems!
pandas is a Python library that makes it easy to read, export and work with relational data. This lesson will expand on its functionality and usage. We typically import pandas as pd to refer to the library using the abbreviated…
Data Science Crash Course
Understanding Probabilistic Models starting with Logistic Regression
Earlier this month, Edward Qian and I started working on a set of comprehensive lessons for aspiring Data Scientists, which can be found on our website www.dscrashcourse.com
I will be cross-posting slightly modified lessons to Medium to make them available to a broader audience. If you find these articles helpful, check out the site for more lessons and practice problems!
Logistic regression is used to model the probability of an event occurring by estimating its log odds. …
Data Science Crash Course
Understanding Tree-based models for Classification and Regression
Earlier this month, Edward Qian and I started working on a set of comprehensive lessons for aspiring Data Scientists, which can be found on our website www.dscrashcourse.com
I will be cross-posting slightly modified lessons to Medium to make them available to a broader audience. If you find these articles helpful, check out the site for more lessons and practice problems!
Tree-based methods can be used for both regression and classification. The idea behind these algorithms is to divide the predictor space into multiple non-overlapping regions and to assign new observations to their respective region. …
Mandy Gu
Data Scientist @ Wealthsimple | Check out my website for learning Data Science: https://www.dscrashcourse.com/
