ML Performance Engineer at Playground

At Playground, we combine AI research and product design to invent new kinds of creative tools.

Note: We are only accepting candidates from the U.S. or Canada at this time

Over the next 10 years, we plan to create a powerful AI render pipeline that allows us to edit, create, and understand pixels to accomplish challenging tasks ranging from instructing an AI to make subtle changes to images, creating entire 3D environments, and understanding context of a scene in a video. We plan to start by accomplishing difficult tasks with images first.

Our mission is to awaken the imagination within each of us.

At Playground, we are making a superhuman AI designer.

Our immediate goal is to build the designer of the future where humans and machines work together to exceed the capability of any single designer. It has exceptional taste, it never tires, it is your partner, it has an eye for details, and hopefully it designs something you couldn’t have imagined yourself.

Read about our latest advancement: https://playground.com/pg-v3

If you join us, you’ll be an early team member in helping shape:

• Our future company culture
• Our engineering practices
• People that we hire
• The direction & focus of our products We need help optimizing training runs to squeeze out the most GPU compute per dollar as well as speeding up inference for our models to keep costs low while providing users a fast experience in the product.

Engineers on the team today:

• Work primarily in Python and PyTorch

• Write code based on the latest research to learn and explore

• A perseverance to experiment but understand not everything they accomplish will be used in our products

• Are supportive—especially when teammates are faced with new challenges

• Are left to autonomously figure out the solutions to their challenges

• Value clear, frequent communication (we do a lot of reading & writing)

• Are naturally curious and willing to take a step to learn something they don’t have experience in

• Feel a great sense of accountability to each other

• Uphold best practices in engineering, security, and design You might be the wrong fit if…

• Publishing papers and earning citations is one of your top 3 priorities

• Writing production worthy code is something you tend to avoid

• Teaching or mentoring people from time-to-time isn’t how you want to spend your time Skills & Experience

• Background in speeding up training runs, optimizing AI models to be faster at inference, etc.

• Track record of optimization/performance related projects

• Having a PhD is not important to us but a track record of building real things in production is

• 4+ years of working full-time Here are examples of things we’ve worked on:

• Building an aesthetic classifier to predict aesthetically pleasing images by user ratings

• Predicting highly accurate object masks from bounding boxes