EarthSnap: Transforming AI Image Identification with MLOps & Managed AI

Earth.com is a premier internet destination for users who are concerned about the planet and the environment and want to make a difference. The company developed EarthSnap, a first-of-its-kind application that enables users to identify all types of plant and animal species via a mobile phone camera. The application is powered by a custom-built, patent-pending AI machine learning solution, which identifies the subject and shares details like habitat, global population distribution, and known history of earth. Founded in 2016 by Eric Ralls, Earth.com Inc. is headquartered in Telluride, CO. The company is committed to providing a vast trove of information regarding life on our planet and functioning as a reference library to assist in research projects.

Earth.com, a leading internet platform for environmental enthusiasts, aimed to accelerate the development and delivery of EarthSnap, an AI-powered image identification application. The goal was to modernize and automate the application’s machine learning (ML) infrastructure, simplify the deployment of new models, and reduce administrative costs. The company insisted on following best practices for end-to-end ML, DevOps, and app development. However, Earth.com lacked an in-house ML engineering team, which made it challenging to add new datasets, improve existing models, release new ones, and scale the ML solution. The models delivered by their previous partner were satisfactory in terms of accuracy but required manual sequential execution for data processing and model retraining. The deployment of endpoints also had to be done manually. Earth.com sought a new strategic partner to streamline the delivery of EarthSnap to market, and Provectus, an AWS Premier Consulting Partner, was chosen for the role.

Provectus addressed Earth.com’s challenges through a series of engagements. They evaluated the work done by the previous partner by investigating data, reproducing ML models, and examining EarthSnap’s backend components. Provectus built fully automated, end-to-end ML pipelines to facilitate the deployment of new model versions, following best practices for data, ML/MLOps, and CI/CD. They also initiated the provision of Managed AI Services, which includes ML infrastructure maintenance, model monitoring and retraining, data quality monitoring, and troubleshooting. Provectus automated the training and deployment of the models, enabling the release of new models without technical support. They also worked on a candidate model, which resulted in success. Provectus offered to enhance the solution with its Managed MLOps Platform, delivered as part of Managed AI Services. The offering included maintenance and support of the existing ML infrastructure, implementation of components for access management, cost monitoring, networking, and compliance support, and 24/7 troubleshooting.

• Provectus helped Earth.com to get EarthSnap back on track. In three months, they designed, built, and implemented fully automated, end-to-end ML pipelines for real-time CV models. The pipelines made ML workflows more observable and facilitated the deployment of new versions of the models, gradually improving and extending them by adding recognition of new species. The enhancements helped Earth.com to reduce engineering heavy lifting and minimize administrative costs. Now, with all components of EarthSnap modernized, Earth.com is ready for future expansion. As part of Managed AI Services, Provectus reduced the infrastructure management overhead, established well-defined SLA and processes, ensured 24/7 coverage and support, and increased overall infrastructure stability, including production workloads and critical releases. They initiated a series of enhancements to deliver their Managed MLOps Platform and improve ML engineering. By collaborating with Provectus, Earth.com was able to release the EarthSnap application on both the Appstore and Playstore ahead of schedule.

• Fully automated, end-to-end ML pipelines delivered in 3 months

• Flexible and easily customizable ML solution for faster, at-scale ML

• Reduced heavy lifting and minimized administrative costs project-wide