Enabling More Efficient Nature Identification with an AI-powered Application

Earth.com is the leading online destination for people who care about the planet. As part of its mission to catalog life on Earth, the company built EarthSnap. The AI-powered mobile app identifies plant and animal species through a phone camera. A user snaps a photo and the app returns the species name along with habitat and natural history. EarthSnap now identifies over 2 million species.

01 The Challenge

A working AI model that needed production infrastructure to become a real product

The species identification app market has grown rapidly. Google Lens, Apple Visual Look Up, and community platforms like iNaturalist all offer automated recognition. For Earth.com, the window to launch depended on how quickly it could move from prototype to product.

EarthSnap’s image classification models performed well. They could identify species from photos with the accuracy the team needed. But the models lived in Jupyter notebooks that required manual, sequential execution. Adding new species, retraining, or deploying updates meant hands-on engineering work at every step.

Earth.com did not have an in-house ML engineering team. The company needed model updates to be simple enough for non-technical staff to manage. The vision was specific: releasing an improved model should take a single click. The team also wanted to reduce infrastructure costs so they could focus on growing the product.

Earth.com partnered with Provectus, an AI-first systems integrator and solutions provider, to automate model delivery and provide support.

02 The Approach

Assess what exists, automate the pipelines, then hand over operations

Provectus structured the engagement in three phases, moving from assessment through automation to managed operations.

The first phase started with understanding what had been built. Provectus reviewed the codebase, cataloged the notebook scripts, and set up an evaluation framework to reproduce prior results. The team then retrained the model on extended data with new species classes, preparing it for production.

The second phase replaced the manual workflow with automated ML pipelines. Instead of running notebooks by hand, the new pipelines handled data preparation, training, evaluation, and deployment automatically. The result: any Earth.com employee could deploy a new model version with one click.

The third phase introduced managed AI services to keep the infrastructure running. This covered ML infrastructure maintenance, cost monitoring, data and model quality checks, retraining workflows, and 24/7 troubleshooting. Provectus also handled the mobile app and backend components, applying CI/CD best practices for public release.

03 The Build

Automated ML pipelines, one-click deployment, and managed operations for a consumer app

The build replaced EarthSnap’s manual notebook workflow with two automated ML pipelines. One handles model training and evaluation. The other handles deployment and release. Together, they let Earth.com improve the model, add species, and push updates without engineering bottlenecks.

The pipelines include:

• Automated data preparation and model training with extended species datasets
• Evaluation frameworks that validate accuracy before any release reaches users
• CI/CD integration so new model versions deploy with a single click
• Cost monitoring to keep infrastructure spend predictable as the dataset grows

The managed services layer runs underneath: infrastructure health monitoring, model quality tracking, retraining workflows, and documentation. The mobile app and backend were also built with production-grade CI/CD. The full product was ready for the App Store and Google Play.

04 The Results

From prototype notebooks to a live app on iOS and Android, ahead of schedule

Provectus completed the full engagement in three months. EarthSnap launched on both the App Store and Google Play ahead of the planned timeline. The team credits the speed to the pipeline automation work.

3 Months

From engagement start to production-ready app

With automated ML pipelines

The operational change was immediate. Model updates that previously required engineering involvement now happen through a single-click deployment. Earth.com can add new species, retrain the model, and push updates without waiting for technical support. The team now focuses on expanding species coverage and growing the user community.

With managed AI services in place, infrastructure runs on defined SLAs with 24/7 monitoring. Administrative costs dropped as the team stopped spending time on maintenance. The company invests in product growth, knowing the ML foundation is stable and scaling with usage.

05 What’s Next

An app that started with species identification and is growing into a biodiversity platform

EarthSnap now identifies over 2 million species. Users share discoveries through an interactive SnapMap. The automated ML pipelines make it possible to expand coverage as new species data comes in. Provectus works with Earth.com on extending EarthSnap’s AI capabilities as the platform grows.