Next Generation Machine Learning Platform
Image Processing Using Machine Learning
Leia Inc. is the leading provider of Lightfield hardware and content services. Lightfield is a new visual medium that transforms existing device displays with lighting effects, texture and 3D depth. It creates a richer experience, making content more beautiful and engaging.
- Founded in 2013 by David Fattal, Pierre-Emmanuel Evreux and Zhen Peng
- Midsize Company
- Clients include Mobile, Automotive, Medical, Retail and Education
Leia is a provider of Lightfield displays and devices and as a startup they were looking for ways to accelerate machine learning setup and experiments. They already had experience with machine learning environments and the Weights & Biases toolset. They were looking for a way to spend more time doing model building and research, and less time focused on getting infrastructure to run. They wanted something flexible and easy to use so that any team member could get started doing training.
Leia conducted market research into the available machine learning platforms, testing out Spell and a few other platforms. After demoing and evaluating environments for about 2 months, they decided on using Spell. The close integration with Weights & Biases (tools for deep learning) was particularly useful for them.
The company already had experience with machine learning platforms and using commands to run Cloud-based scripts. The engineers found it easy to start using Spell and quickly ramped up on the command language without needing extra time for training. Within a week they were able to get new engineers up to speed on Spell.
Edward Li, an engineer at Leia, noted “The uptime to bringing on a new engineer is literally less than a week, and I think that’s something that’s a very big pro for your platform.”
Spell has provided a very easy way to run experiments, hyperparameter searches and concurrent jobs. Leia is able to manage the resource usage of their clusters more easily so they don’t have wasted resources or surprises in billing at the end of the month.
The features that have made the biggest impact on their day-to-day workflow have been GCP storage configuration, defining machine types, and the ability to easily run jobs in parallel. Simple, straightforward UI has made it easy to use.
Owen Hua, computer vision engineer, noted "We are not only using Spell, as we also have local machines, but we've found Spell gives us the maximum flexibility to explore different experiments and ideas we have."
Leia is rapidly growing and anticipates using Spell will help with their growth because its ease of use makes it quick to onboard new engineers. Since it uses their existing Google Cloud resources, it helps understand their usage and it’s not a magic black box.
“Given that its using our own Google Cloud resources, it helps us understand what’s going on,” said Puneet Kohli, Leia engineering manager. “When we get the bill for Google Cloud it’s very clear that this was the part for machine learning training jobs and we can map that back to Spell and understand the usage very easily.”
"We have been rapidly growing recently, the ease of using Spell allows us to onboard people really easily and run multiple jobs pretty fast."
Machine Learning Projects with Spell
Request a Demo
Schedule an in-depth demonstration with a Spell representative to learn how Spell can help streamline and accelerate your machine learning development.