EarthDaily Analytics
Innovating Change Detection Modeling
During periods of intense experimentation and integration, the researchers involved could spend as much as a day a week resolving dependencies due to diverse tools and frameworks, slowing down their experimentation and prototyping speed. Further, only 40% of the team had access to their on-prem GPUs at a given time. EDA needed a unified experimentation platform to facilitate rapid iteration and leverage best-in-class software, like VS Code, for steady team growth.
“Lightning AI is perfect for experimentation. If you want to do something radically different with your experiment’s development environment, you don’t need to worry about the implications of that for others.”
In just two days, Lightning AI’s transformative solution, Lightning AI Studios, streamlined and simplified EDA’s onboarding, and integration of R&D engineers from culturally diverse teams across two different divisions into a common prototyping platform for a cross-company hackathon. Lightning AI Studios’ turn-key approach minimized challenges related to differences interfacing with GPU hardware while maintaining a customized environment.
This innovation enabled decentralized research, empowering independent movement and rapid iteration without the complexities of merging work or risking dependency jail. Access to on-demand, powerful GPUs through Lightning AI Studios not only provided intangible benefits but also allowed 60% of the ML team within EDA to experiment freely, saving significant time in managing and debugging Docker images during machine switches.
Leveraging their AWS account with Lightning Studios, EDA was able to customize networking and permissions to integrate with their existing ecosystem of tools securely. The team’s agility was further demonstrated by integrating with an existing self-hosted MLFlow instance in just two days, highlighting the platform’s extensibility and preventing the need for a stack overhaul.
Lightning AI Studios’ self-serve capabilities and sharing features enhanced collaboration, enabling researchers to seamlessly share the environment and the Studio while reserving code versioning and management to Gitlab. This fostered a sense of connection within the distributed team. In summary, Lightning AI’s Lightning AI Studios has not only improved the efficiency of ML-based R&D but also empowered the EDA team with newfound agility and collaboration capabilities.
- Studios
- Lightning Trainer
- VS Code integration
- Extended the platform with integration to an existing self-hosted MLFlow instance in just two days
- Enabled a team of engineers from vastly different research teams and from different countries to begin collaborating in only 1 to 2 days
- Increase ML R&D productivity by eliminating up to a day of infra management and dependency resolution, especially during periods of intense experimentation and integration
About EarthDaily Analytics
EarthDaily Analytics is a vertically integrated software and analytics company utilizing cutting-edge Big Data tools and proven Space technologies to provide value-added insights to the people, businesses, and governmental entities confronting the world’s most pressing challenges. The ML R&D team is committed to leveraging daily global satellite imagery to develop models that pinpoint large-scale environmental changes.
With the aid of Lightning AI, EarthDaily Analytics has supercharged collaboration within its R&D team and increased experimentation velocity. They’ve achieved this by streamlining their model re-training pipeline, ensuring its AI-driven analyses remain at the forefront of environmental monitoring.