Determined Hackathon 2020: Remotely Successful!

This past November, Determined AI held its first-ever hackathon. All Determined employees - not just engineers! - took the better part of two days to collaborate, iterate, and challenge each other to build the next product feature of Determined’s deep learning training platform.

Hackathon background

Working remotely since last March has been a unique challenge, so finding ways to remain a tight-knit group is of the utmost importance. In the past few months, we’ve virtually held a company retreat, team poker night, and a “white elephant” gift exchange for the holidays. The hackathon was a chance for folks from different corners of the organization to work with each other on the product and perhaps outside of their normal to-do list.

We thought the only way to judge the hackathon presentations was to have the criteria fall directly in line with Determined AI’s company values:

  • Always be learning - Points for going outside of your comfort zone, or for advancing an idea that is new to the rest of the team.
  • Build the strongest team - The presentations should demonstrate that the team worked well together. We want to see a truly cross-disciplinary effort, with no one person doing the heavy lifting.
  • Customer value - The presentations should clearly show how customer value could be created if this idea were to be explored further.

Individuals sorted themselves into teams by whichever project caught their attention or sparked their interest. The project proposals this year were:

  • Arbitrary Python Models in Determined
  • Webflow for our company website
  • Support for PyTorch & TPUs
  • Sharable Determined Artifacts
  • Model Interpretability
  • Model Spend Executive Report

Hacking

Teams split off into separate Zoom meetings to run experiments, test theories, and assemble slide decks to show off their work.

Remote hacking was a challenge, but here’s a glimpse into the madness:

Hackathon madness

Hackathon madness

Hackathon madness

Hackathon madness

Results

After two days of collaboration, we reconvened for the final presentations. Each team had ten minutes on the clock to talk about the importance of their project, discuss why it was the best fit for a new product feature, and present their findings to the company.

Each team brought something different to the table. Furthermore, not every project was centered around making Determined’s deep learning training platform better! Some ideas were geared towards making the product appealing to higher-ups with decision-making power, some were based on making Determined employees’ lives easier.

Once the presentations were finished, we took to Slack where everybody individually voted on their favorite project:

The votes are in

In a tight, hotly-contested race, taking home the crown for this hackathon was team Arbitrary Python Models. We’re looking forward to the implementation of their hacking in an upcoming release of Determined. This feature will allow users to access basic Determined features - like experiment management, resource management, hyperparameter tuning, and distributed training - on arbitrary ML/DL frameworks or libraries that have built-in training support.

But wait - there’s more! After the hackathon, most of the team stayed on for a remote Escape-The-Room, collaborating to track down clues and break out of a (virtual) maze. Zoom fatigue is definitely a real thing, but that didn’t stop our team from having fun.

What’s Next?

Determined’s 2020 Hackathon was a success! We’re constantly getting feedback from our open-source community on which features and functionalities to prioritize next, but seeing Determined team members working hard on projects to test the limits of the product was incredibly valuable. We look forward to collaborating on another hackathon in 2021.


We’re hiring, come join us!

Does this sound like the type of project you’d love to work on, and do you have hands-on experience with deep learning, ML infrastructure, or systems? We’d love to have you join the Determined team! We’re currently hiring for Front-End, Support, and Systems Engineers, as well as ML Developer Advocates. Check our careers page for more information.