Create Scoring Data
Scoring data needed to train a custom AI Scoring Algorithm.
Scoring algorithms
A scoring algorithm serves as an AI-powered tool for AI matching. It's directly deployable to create recruiter or talent copilots in the AI studio.
1. Standard Scoring Algorithms:
Our default scoring algorithms are rigorously trained on diverse datasets to ensure :
- 💪High performance: Human level matching capabilities at low cost.
- 🎯Objectivity: Minimal Bias for maximal fairness considerations
- 🔄Adaptability: Easily adjusts to match specific organizational needs and preferences.
2. Fine-Tuning for Optimal Performance:
Every company is unique, so we've implemented algorithm fine-tuning. This process allows you to adjust and improve our algorithms based on your specific dataset distribution.
Fine-tuning creates an enhanced, private version of the default algorithms, capturing the nuances of each organization. The trainable algorithms have the following features :
- 🎛️Fine-tuning: Easily retrainable with low quatities of data.
- 🧠Minimal knowledge loss: Fine-tuning doesn't decrease the previous knowledge of the default scoring algorithms.
3. Viewpoints:
Scoring algorithms are designed to function for multiple audiences. Each audience has its own viewpoint and objectives:
Audiences | Definitions | Objective | Event Types | Connectors |
---|---|---|---|---|
Recruiters | People responsible for searching and interviewing both candidates and employees for positions | Recruit the best candidates | ATS stages, HCM stages, Interviews ... | Pixels, ATS, HCM, CRM, API |
Candidates | Job seekers navigating job or career sites | Find the best job | Job views, Apply buttons, Form submissions, Email clicks ... | Pixels, Heatmaps |
Employees | Individuals looking to transition to a different role within their organization | Find the best role | Training completions, Performance reviews, Meeting attendances ... | Pixels, App Analytics, API, ETL |
Updated 5 months ago
Depending on your specific use case, please read the following articles: