The learning curve for machine learning is way too steep. As a result, non-data scientist developers avoid building products that depend on machine learning algorithms. Telepath is changing all of that.
Upload your data (in almost any format)
Answer non-scientific/non-statistical questions about your data
We automatically determine the best type of machine learning algorithm to use and build you a predictive model
You get a developer-friendly API endpoint
Send data to your model and get back instant predictions
We automatically monitor and tune your model to keep it accurate
Let's assume that you want to build a churn-detection feature for your SaaS startup (or maybe you want to build a tool to sell to other startups). Here's how Telepath can help!
Gather the data that you have about your user activity, such as page views, button clicks, feature interactions - the same kind of events that you'd send to Segment or Mixpanel.
Once you upload your data, we'll ask you some questions: What's an unusually low and high value for "Column ______?" "Are the values of Column ____ strings or numbers?" "Which column are you trying to predict (which column represents when a user has churned?)"
The Telepath AI platform runs your data through a long list of machine learning processes to determine the most predictive type of process and algorithm to use for your project (i.e. is this a linear regression problem? a neural network?). This process is sometimes called "automatic machine learning" or Auto ML.
We deploy a production-ready model into the cloud and give you an API endpoint to generate your predictions. The whole thing is fully-managed for you.
Now you're ready to use your model to make predictions. When a user generates new activity (clicks on another button, uses a feature, etc) you send that data to your Telepath API and instantly get back a prediction on how likely that user is to churn. Whenever a user does churn, you also send us that data, and we constantly re-train the model to improve its accuracy.
Want to join our private beta?