This page is taken un-edited from our internal wiki. If some statements appear oddly worded, its because it’s how we speak internally to our own team-mates.

V7’s remit is to help users create AI, improve AI, and understand what it knows and doesn’t know. As we know AI’s input and output is in labels. V7 is in fact a platform to overlay labels (AI’s knowledge) on top of raw data for both training and inference purposes.

What V7 is not:

• A platform to run training jobs in high detail
• A platform for running AI on the edge

3 steps to V7’s success:

Step 1: Develop a great labeling experience

V7 should be the place where training data and prompts can be easily manipulated to influence the output of AI.

Step 2: Develop workflows to orchestrate models and humans for labeling

This allows us to programmatically add or edit what AI knows at massive scale. We move away from using tweezers to improve AI to creating an “immune system”

Step 3: Let customers tune foundation models to pre-solve any task.

V7 pre-completes tasks, pre-identifies issues with your models, and helps fully automate processes using general-purpose AI models. The immune system now gains a nervous system to bring in agency.

V7 is a platform built to help machine learning teams create and run more accurate AI models by feeding them better data.

Our Customers find that by using V7, their error rate in their training data sets is reduced by 30-50%