NeoPulse Manager

Data Preparation

The NeoPulse Platform can accept training datasets from different sources.
Using the in-built data viewer, the user can inspect the data, and identify the input and output columns.
The dataset can be archived for auditing purposes.

How to add a view in NeoPulse Manger

Creating & Importing AI Models

The visual model builder will solve most problems automatically with an auto block
One can use the visual editor to drag and drop layers and get help connecting them
Advanced users can write NML code directly and inspect it in the visual model builder
Import AI models created with Keras, Tensorflow, Pytorch or Chainer
Import pretrained models for retraining

Training & Evaluation of AI Models

Users can choose a range of loss functions and metrics for evaluating models
Users can choose data for validating models or let NeoPulse® choose
NeoPulse® explainability features allow users to open ‘black box’ AI models and answer the question ‘why?’
Using NeoPulse Manager, a user can evaluate the best model for deployment with a couple of clicks
Deployment is simply a drag and drop operation – dragging a model to the right NeoPulse® Runtime cluster or NeoPulse® Runtime group will distribute the model to all the nodes in the group or cluster automatically

How to deploy a PIM with NeoPulse Manger
NeoPulse Manager DevOps

NeoPulse DevOps Features

Monitor AI models in production
Stage AI models for inference
Group deployment targets for easier management
Perform A/B testing to replace models in production
Verify status of production environments

«
View NeoPulse Manager Demo
»

NeoPulse can solve hundreds of real-world problems across dozens of industries: industry prediction, preventive maintenance, protein synthesis, repurpose drugs, detecting threats, and much more! Customers can build state-of-the-art models in minutes unlocking millions of dollars of NPV for their businesses every year.

In this live demo, NeoPulse Manager built a state-of-the-art machine learning model using COVID data from Kaggle with very minimal effort, beating the highest model currently listed with an accuracy of 95.27.