Run Python scripts on cloud GPUs

One command to execute any Python script on NVIDIA GPUs. Zero-setup cloud GPU computing with automatic provisioning.

42x Faster Training

PyTorch CIFAR10 Training Time

Local CPU

400 mins

Apple M1 GPU

85 mins

NVIDIA T4 GPU

9.8 mins

Cost: $0.12 • 42x faster than CPU

GPU-accelerated Python, made simple

From local Python script to cloud GPU execution in one command.

Any GPU, Any Cloud

Tesla T4s, A100s, H100s on AWS, GCP, and Azure

PyTorch, TensorFlow, OpenCV, Numba, CuPy, etc.

Auto-shutdown when done (no idle costs)

Zero Setup

Auto-sync local Python packages, pre-configured with NVIDIA drivers

Any custom deps you want (no environment restrictions)

No K8s or Terraform, Docker optional

Managed GPU Compute

Autoscaling to hundreds of GPUs just when you need them

No need to configure job queues, compute environments, or IAM roles

Monitor GPU utilization, logs, and hardware metrics all in one place

From your laptop to cloud GPUs

Run a Python script on a GPU-enabled cloud machine with one line of code.

Easy to get started, extend as needed.

Take any Python script and run it on cloud GPUs
Scale out with --n-tasks for parallel processing
Specify instance types with --vm-type
Use custom Docker containers with --container

See Full Example

$ coiled batch run --gpu python train.py

$ coiled batch run --gpu python train.py

From serverless GPUs to large-scale, long-running batch computing

Choose the tool that fits your workflow.

Serverless Python Functions

Serverless GPU computing: Add a decorator, run on cloud GPUs.

import coiled
import torch

@coiled.function(
    vm_type="g5.xlarge",     # A10G GPU
    keepalive="20 minutes",  # Warm starts
    region="us-west-2",      # Any region
)
def train():
    ...
    return model.to("cpu")

model = train()

import coiled
import torch

@coiled.function(
    vm_type="g5.xlarge",     # A10G GPU
    keepalive="20 minutes",  # Warm starts
    region="us-west-2",      # Any region
)
def train():
    ...
    return model.to("cpu")

model = train()

Automatic environment syncing (even CUDA drivers)

Scale out for parallel model tuning and inference with .map()

Perfect for model training and interactive work

See Serverless GPU Example

Batch Jobs

Embarrassingly parallel batch jobs on cloud GPUs.

#!/usr/bin/env bash

# COILED n-tasks 10
# COILED gpu True

accelerate launch \
    --multi_gpu \
    --machine_rank $COILED_BATCH_TASK_ID \
    --main_process_ip $COILED_BATCH_SCHEDULER_ADDRESS \
    --main_process_port 12345 \
    --num_machines $COILED_BATCH_TASK_COUNT \
    --num_processes $COILED_BATCH_TASK_COUNT \
    nlp_example.py

#!/usr/bin/env bash

# COILED n-tasks 10
# COILED gpu True

accelerate launch \
    --multi_gpu \
    --machine_rank $COILED_BATCH_TASK_ID \
    --main_process_ip $COILED_BATCH_SCHEDULER_ADDRESS \
    --main_process_port 12345 \
    --num_machines $COILED_BATCH_TASK_COUNT \
    --num_processes $COILED_BATCH_TASK_COUNT \
    nlp_example.py

Run existing scripts with minimal code changes

Auto-provisions and configures GPU VMs

Perfect for image processing, distributed training, and batch inference

See Batch GPU Example

Trusted by Data Teams

Reliable GPU compute for mission-critical workloads

"I've been incredibly impressed with Coiled; it's quite literally the only piece of our entire ETL architecture that I never have to worry about."

Bobby George

Co-founder, Kestrel

"The speed is nice, sure, but the real benefit is taking a multi-day effort and finishing it in an afternoon. Coiled changed the character of our work."

Matt Plough

Software Engineer, KoBold Metals

"My team has started using Coiled this week. Got us up and running with clusters for ad hoc distributed workloads in no time."

Mike Bell

Data Scientist, Titan

"Coiled is natural and fun to use. It's Pythonic."

Lucas Gabriel Balista

Data Science Lead, Online Applications

FAQ

AWS Lambda doesn't support GPUs, but Coiled does.

If you're looking for "AWS Lambda with GPU support," Coiled is what you need:

Annotate your Python functions: @coiled.function(gpu=True)
Auto-provisioning: Spins up GPU instances automatically (like Lambda for CPUs)
Zero infrastructure management: No servers to manage, just like Lambda

Perfect for ML inference, training, and any GPU-accelerated Python workload.

You can use any Python library with Coiled.

Some popular GPU-accelerated libraries:

PyTorch: Automatic CUDA version matching
TensorFlow: GPU-enabled by default
OpenCV: GPU-accelerated computer vision
CuPy: NumPy-like GPU computing
Numba: CUDA kernel compilation
Rapids: GPU-accelerated data science

Your Python code automatically detects and uses available GPUs.

Example GPU detection:

import torch

# PyTorch automatically uses GPU if available
if torch.cuda.is_available():
    device = torch.device("cuda")
    print(f"Using GPU: {torch.cuda.get_device_name()}")
else:
    device = torch.device("cpu")

import torch

# PyTorch automatically uses GPU if available
if torch.cuda.is_available():
    device = torch.device("cuda")
    print(f"Using GPU: {torch.cuda.get_device_name()}")
else:
    device = torch.device("cpu")

Coiled ensures CUDA drivers and libraries are properly configured.

Yes! Coiled Functions provide true serverless GPU computing.

With @coiled.function decorator:

@coiled.function(vm_type="g5.xlarge")  # GPU instance
def process_image(image_data):
    # Your GPU code here (PyTorch, OpenCV, etc.)
    return processed_result

@coiled.function(vm_type="g5.xlarge")  # GPU instance
def process_image(image_data):
    # Your GPU code here (PyTorch, OpenCV, etc.)
    return processed_result

Learn more.

Ready to run Python on GPUs?

Get started in under 2 minutes. Your first 500 CPU hours per month are free.

See Documentation

$ pip install coiled
$ coiled quickstart

Grant cloud access? (Y/n): Y

... Configuring  ...

You're ready to go. 🎉

$ pip install coiled
$ coiled quickstart

Grant cloud access? (Y/n): Y

... Configuring  ...

You're ready to go. 🎉