Pachama Uses Flyte to Calculate and Sell Carbon Credits at Scale
The company
Founded in 2018, Pachama has helped dozens of Fortune 500 companies invest in high-quality, nature-based carbon credits.
Pachama’s mission is to restore nature to solve climate change. The San Francisco company uses data, artificial intelligence and automation to protect ecosystems and forests and improve the carbon market.
Pachama is founded on the belief that properly quantified land has a much larger value — it’s important to look at the entire economic value of land that would typically be deforested or otherwise used with adverse effect. The most tangible economic value is carbon capture and storage.
The carbon credits Pachama sells comprise a financial instrument that compensates landowners to leave forests intact, helping the surrounding ecosystem and the Earth’s climate. (“In a way, we are a marketplace, like Airbnb, connecting supply and demand,” Pachama Co-founder and CEO Diego Saez Gil remarked during a May 2021 panel discussion on NPR’s “Planet Money.”)
The challenge
The carbon credit is highly dependent on the integrity of the measurement on which it’s based. To quantify the asset, Pachama needs to quantify the carbon value of the land. An improper measurement can have a net-negative impact.
The measurement of the carbon credit is the change in carbon from the project or intervention versus the change in carbon without the project being established. Trying to figure out what would have happened is the hardest part.
The solution
Pachama’s assessments employ algorithms to calculate the impartial baseline of credit numbers. By harnessing remote sensing and machine learning, Pachama creates carbon maps of the areas to be analyzed which helps increase transparency, quality and scalability in the carbon market. By using Flyte orchestration, the team is able to use biomass metrics and modeling at scale.
With the map data in hand, Pachama uses Flyte to perform a baseline calculation of the carbon value of the land and to create biomass modeling, both which go into calculating carbon credits.
Bernhard Stadlbauer, Data Engineer at Pachama, identified a few ways Flyte’s platform has been integral to their success:
Reusable workflows make it easy to share complex ML tasks among teams without having to assign all dependencies to one container.
“Flyte workflows don’t break when scaling,” Stadlbauer said. “Before Flyte, we were using ad hoc code. Anytime someone wanted to do something in a similar way, they had to ask around and copy and paste.
“Flyte is good at keeping the code and all of the things in between. Our productivity has improved and we can run more experiments and do more. As our central components have matured, Flyte’s become a no-brainer.”
Pachama also uses Flyte’s fast registration to find Docker images each time that are similar to the current code. The Pachama team created its workaround for the problem by writing a small script that traverses the git graph and checks against the registry until it finds the Docker image that’s closest to the current code, which it uses for registering. The process also goes back to the main branch to ensure that the correct image is selected.
Caching makes debugging very easy — it reduces the duration cycle because work doesn’t have to be redone.
“If a task fails, it’s usually a 1-1 mapping to local workflow,” Stadlbauer said. To simplify and speed debugging, Stadlbauer downloads the data and debugs locally.
The Flyte community provides constant feedback and support; Stadlbauer said it was the single most important benefit of adopting Flyte.
“I’ve worked with Flyte for almost two years, and we’ve gotten a lot back,” he said. “We’ve gotten so much support.”