Machine Learning Services
Praxyk is a 100% open-source Machine-Learning-as-a-Service provider.
- Choose A Model - Praxyk offers a variety of pre-trained machine learning models for you to choose from. Everything from optical character recognition, to facial detection, to spam detection models are pre-trained and live on our servers.
- Upload Your Data - Upload anything from a few bytes to terabytes of data directly to our servers.
- Receive Predictions - We'll run your data through the chosen model on our fast, stable, and secure architecture and return the results to you.
Praxyk plans to offer two main types of services, Prediction on Demand and the Templated Learning Platform. As of right now only the former is implemented, but we will keep you updated with our plans towards the latter. These two services are described below.
Prediction on Demand
Prediction on Demand (POD) is a way for you to access world-class and pre-trained machine learning models instantly through our API. We have chosen some of the most common use-cases of machine learning, facial recognition for instance, and developed powerful models to solve these common problems. We then host these models on our server farm and offer access to them via our API and website. Some of the models offered by POD are given below.
- Optical Character Recognition
- Facial Recignition
- Spam Detection
Templated Learning Platform
The templated learning platform is the natural extension to the POD service. Instead of hosting static models for the public to use, the templated learning platform allows you to start with templates of machine learning models which can then be trained to suit your more specific goals.
We strive to accomplish two primary goals :
- Abstract the issue of designing and maintaining a custom Machine-Learning framework away from the customer.
- Make the predictive power of Machine-Learning more accessible, especially for individuals and small businesses.
This is a project for CS115 Intro to Software Engineering at UCSC, Fall quarter 2015.
- John Allard
- Nicholas Corgan
- Nikita Sokolnikov
- Ryan Coley
- Nick Church
- Michael Vincent
This project followed Scrum methodology; relevant Scrum documents can be found in the docs directory.