Hrishi's Blog

A better approach

Training a text-classifier “AI model” which takes an input string and spits out structured output that looks something like:

 {
    jobTitle: "software engineer",

    jobTitleId: 456,

    location: "San Francisco, CA",

    locationId: 335555,

    confidence: 0.87
 }

The text classifier would be trained on a large amount of text, so the model “learns” what terms are probably job titles, and what terms are probably locations. This model can then be queried by the server side code by passing it the string “software-engineer-jobs-in-san-francisco”. When the server gets back the structured output, it can do a straightforward query to it’s job postings database, and return relevant responses