The API for ML Compute

The fastest and lowest cost ML compute service in the world.

Only pay when you train, predict and deploy.

import npu
npu.api(TOKEN)
# Preprocess and pack your data
...
# Create model architecture (Pytorch, TF2 or MXNET)
...
# Upload model to API
model = npu.compile(your_model)
# Train your model with your data
trained_model = npu.train(model, train_data, ...)

The problem with current infrastructure

Cloud bills get expensive and uncontrollable. On average 25% of revenue from AI companies is spend on cloud resources. On top of this cost, every AI company is required to build their own scalable and optimised infrastructure to train and deploy their AI models.

We wanted to help, so we built the Neuro API.

About the API

We’ve created a dead simple API that enables data scientists and AI teams to prototype and deploy ML workloads in top-tier cloud hardware in seconds. We’ve built the infrastructure so you can focus on the AI.

Only pay when you train or predict. No extra over-heads and control exactly how much you spend.

How it works

  1. Build your model with your favourite AI framework. 
  2. Upload you model and dataset via the API.
  3. We use ML to distribute your tasks on the best hardware and cloud provider.
  4. View all your tasks running in the Dashboard.
  5. Get the trained model or results after the task is completed.

Stay in touch with us