CASE STUDY

X

Increasing Collaboration, Efficiency & Productivity

  • Founded in 2007 by Jeff Scheinrock, CFO and President
  • Midsize company
  • Based in San Francisco
  • 150+ clients, including National Geographic and McKinsey and Co.

OVERVIEW

Originate is a concept-to-launch digital product company providing a breadth of services that accelerate innovation and improve clients’ processes. The company conducts extensive research to inform their services, enabling clients to modernize seamlessly and remain at the vanguard of their respective industries.

BACKGROUND

One of Originate’s cutting-edge product offerings is the Machine Learning Cloud in their Origin product, which provides analytics, modeling, and inference tools to clients. However, without the help of a powerful platform, the reality of experimental ML projects is hours handling data instead of training models and iterating. 


Amanda Doucette and Danny Meyer, two machine learning engineers at Originate, work in tandem on research-based projects requiring comprehensive data analysis, experimentation, and collaboration. Before Spell, they shared a single machine for ML experiments, which hurt their ability to iterate efficiently on experiments. In addition, they lacked effective collaboration tools, meaning reviewing each others’ runs meant physically finding the other person in the office. This process cramped productivity and their ability to onboard new clients. 


“If you have a client project, I don’t know a faster way to get up and running than Spell.”

Danny Meyer, Software Engineer

SOLUTION

At the suggestion of a colleague, Amanda and Danny implemented Spell, and it rapidly transformed the Originate team workflow. Where they once spent hours worrying about DevOps, the team now moves seamlessly from prototype to production, running experiments on Spell’s ready-made platform with multiple GPU machine types. And on top of that, managing their data housed in Amazon Cloud Storage’s S3 buckets has been a breeze with Spell’s file system.


With Spell’s collaboration tools, the team stays connected and in control of their runs, regardless of whether they’re in the office. Engineers easily monitor tasks, share progress, and enjoy the boost in creativity fueled by an increase in visibility. “We love the ability to see what’s going on,” said Danny, noting that the more information they can conveniently share, the more the team can learn from one another. 

RESULTS

Within months time, the team trained hundreds of models on Spell, all the while bringing compute costs down 65%.

 

Their new and improved workflow on Spell makes machine management effortless and cost-effective: Spell automatically spins down machines once they’re no longer needed and provides access to Spot Instances. With Spot Instances, the team saves money without ever worrying about losing their work, which stays saved through the auto-backup feature. 

The team works better together, and is further augmented by the reliable support team at Spell. Danny raves, “The people at Spell have been the best. When there’s an issue they’re extremely helpful and hop on to get it taken care of.” Whenever problems arise, Originate engineers know they can lean on Spell Support. 

 

Spell has been instrumental in boosting productivity and revenue, as well as eliminating previous operational limitations on the Originate team. Data and machines are in one central, secure location, and the team is empowered to try new approaches, spin up multiple machines, and deploy models with ease. “If you have a client project, I don’t know a faster way to get up and running than Spell,” says Danny. 

Spell is a powerful platform for building and managing machine learning projects. Spell takes care of infrastructure, making machine learning projects easier to start, faster to get results, more organized and safer than managing infrastructure on your own.


contact@spell.ml