USA Virtual Hackathon | July 30 - August 24 | 18+ Event

Registration Capacity Full


Congratulations to the 2020 Environmental Virtual Hackathon Winners!
1st Place: VirtEnvHackathon
2nd Place: SenseAnomali
3rd Place: Autoencoder
Would you trade in your office for a four-story sphere surrounded by 40,000 plants? The Spheres are a result of innovative thinking about the character of a workplace and an extended conversation about what is typically missing from urban offices – a direct link to nature.
AWS and NVIDIA are hosting a hackathon where you can build your own solutions around helping the environment. Whether you are passionate about stopping deforestation, tackling climate change, or are also interested in bringing nature into urban environments, we want to see what you can do.
You bring the big ideas and we’ll provide the AI tools and powerful cloud infrastructure for you to build your vision.
AWS’s Seattle Spheres
The spheres range from three to four stories tall, house 40,000 plants as well as meeting space and retail stores. The domes are kept at a temperature of 72 °F (22 °C) and 60 percent humidity during the daytime with sensors that collect data about temperature, humidity, light, CO2, dew point, and other environmental variables. They were created to not only provide a creative workspace for AWS employees but to be an example of how to think big about reintroducing nature into urban environments.
Learn More About the Spheres



  • Cash Prize: $ 15,000.00


  • Cash Prize: $ 5,000.00


  • Cash Prize: $ 3,000.00

The Challenge

Build, train, optimize, and deploy a machine learning model to a virtual edge device that accurately detects anomalies

Example Topics

Soil and Land Pollution

Loss of Biodiversity

Agricultural Pollution

Climate Change


Sample Training Data

Sample Data – Single Day

Sample Data – 3 Consecutive Days

Sample Data - Day 1
Sample Data - Day 1
Sample Data - Day 1

Each of the 23 sensors is sampled at 15 minute intervals, therefore a day has 96 values per sensor concatenated with its sensor peers in the following order:

sensorList = ( ‘co2_1′,’co2_2’, ‘co2_3’, ‘co2_4’,
‘temp_1’, ‘temp_2’, ‘temp_3’, ‘temp_4’,
‘dew_1′,’dew_2’, ‘dew_3’, ‘dew_4’,
‘instLight_1’, ‘instLight_2’, ‘instLight_3’,
‘relH_1’, ‘relH_2’, ‘relH_3’, ‘relH_4’,
) + day and hour index

Additional details will be provided at the competition.


Amazon SageMaker

Amazon SageMaker is a fully-managed machine learning platform that enables developers and data scientists to quickly and easily build, train, and deploy machine learning models at any scale. Amazon SageMaker includes the most common machine learning algorithms and is also pre-configured with the latest versions of frameworks such as TensorFlow and Apache MXNet.

Learn more.

Amazon EC2 P3 Instances

With Amazon EC2 P3 instances, powered by the NVIDIA Volta architecture, you can significantly reduce machine learning training times from days to hours.

Learn more.

NVIDIA Platform

NVIDIA GPUs along with the CUDA-X AI libraries are supported on AWS EC2 P & G instances as well as AWS SageMaker for Training & Inference. NVIDIA GPUs are the most widely adopted platform for DL training and the most performant for real-time inference, and accelerate all deep learning frameworks. This complete platform saves data science researchers and developers time and money, letting them bring next-generation AI-powered services to life.

Learn more.

Amazon EC2 G4 Instances

G4 instances are optimized for machine learning application deployments (inference), such as image classification, object detection, recommendation engines, automated speech recognition, and language translation that push the boundary on AI-innovation and latency.

Learn more.


Tutorials & Sample Code

Amazon SageMaker

AWS Deep Learning AMIs

AWS Credits

Each attendee will receive $100 AWS credits to build out their solutions.  You will be given your code once the competition begins on July 30. Make sure to create an AWS Account before.

Judging Criteria

Each submission will be scored in each round based on the following criteria with a minimum score of 0 and maximum score of 20 points, with the final score being the average of the judges’ scores:

Accuracy (10 points):

Is the model accurately identifying anomalies based on input data?

Design/Visualization (5 points):

Was the UX/UI intuitive and appealing?

Simplicity (5 points):

Is the application simple to use and can the team explain it clearly in three sentences or less?


Wenming Ye

Sr. Solutions Architect, AWS

Miro Enev

Sr. Solutions Architect, NVIDIA

An Luo

Sr Technical Program Manager, Academic Programs, AWS


  • Teams of up to 5 participants are allowed. All team members must have completed the participation agreement to compete.
  • Participants must be 18+ years old by the start of the hackathon and a resident of the United States.
  • You may not begin your project until the competition officially begins. Please don’t build on top of previous projects if you want to win.
  • Winning teams will be subject to a code-review at some point following the event or immediately before winning.


Who Can Participate?

All developers, designers, and entrepreneurs who are 18+ and residences of the United States. 

How are Teams Formed?

Teams can be created in advance using the Environmental Hackathon Slack channel and can be comprised of 1 to 5 individuals.

How is the Environmental Hackathon Judged?

Your project will be submitted and judged through: 
Each team will submit a 2-3 minute video demonstrating their submission.

What is the fresh code rule?

All code developed as part of the Environmental Hackathon event must be fresh. Before the start of the hackathon, developers can create wireframes, designs and user flows. They can also come with hardware. But to keep things fair, all code must be written during the official hackathon time period. Other than that, almost anything goes and you can use any coding languages or open-source libraries.