Gaia AI

We are committed to solving the issues of trust in the forest carbon offset market, which is largely driven by a lack of high resolution data that would give an accurate measurement of carbon stocks in a forest.

Our solution is to equip drones with cameras and LiDAR sensors - the same sensors used in autonomous vehicles - and to fly these drones around a forest to gather rich, high resolution data on the forest.  We would then apply similar algorithms to what we have worked with in the autonomous vehicles’ space, to automatically process this high resolution data into the forest metrics and carbon stock estimates that stakeholders need to have high confidence on carbon credits.

If scaled globally, our solution would help fight against climate change by enabling more forest carbon offset (conservation) projects, and it could also provide economic opportunities to rural areas around forest conservation projects.

Recent studies say that even in an optimistic climate scenario where we rapidly reduce our CO2 emissions by 2050, we will still need to capture CO2 at a rate more than 125x the 2016 levels to avoid a climate catastrophe. Nature-based CO2 removal - specifically, planting and growing trees - is a tried and true method of capturing carbon and can be used to generate carbon credits for land owners. However, the current approaches to verifying forest growth (to generate carbon credits) involve massive amounts of manual labor.

This practice is both expensive and slow. In addition to all this, there is a lack of available data for project developers to intelligently design and manage forests. Furthermore, the lack of high resolution data on forests has instilled a lack of trust in forest carbon offset projects, limiting the potential of forest conservation as a way to tackle climate change.

Gaia uses drones, computer vision and LiDAR to measure forests. We are equipping drones with cameras and LiDAR sensors - the same sensors used in autonomous vehicles - and flying these drones around a forest to gather rich, high resolution data on the forest. We will then apply the same algorithms used in the autonomous vehicles space to automatically process this data into actionable insights about the forest, such as number and species of trees, volume of trees, and ultimately a precise estimate of volume of carbon stock in a forest.

Landowners and project developers can then use these insights to better manage their forests, as well as help them to monetize conservation of their forests by participating in the natural carbon offset market (which is where corporations are now investing money to offset their own carbon emissions).

If we succeed, we will help conserve forested land around the world, sequestering more carbon from the atmosphere and therefore helping to fight climate change. Climate change is an issue that affects the entire planet, so our solution will impact the broader population.

More tangibly, our solution will also support foresters in their work, helping them to work more efficiently while also giving them the tools to measure forests more accurately. Also, in many situations - especially for some of the more remote forest carbon offset projects - foresters have to traverse more treacherous environments to manually measure plots of land. By creating a drone data acquisition system, we will help foresters accomplish their jobs more safely.

Finally, by measuring the health of forests, we will enable better management of forest ecosystems, which will improve the lives of communities who live around these forests. These communities tend to be more rural, and much of their livelihood depends on the local environment. Improved forest management can also lead to better food and water supply for these communities, not to mention the jobs that may come from managing and conserving more forested land.

First, we are addressing the problem of efficient, accurate data collection in forests to measure biodiversity and carbon content. Our solution uses sensors and AI similar to autonomous vehicles to collect and process forest data into meaningful metrics.

Furthermore, we use machine learning to classify trees and extract insights about forest layouts, including how resiliency, biodiversity, and carbon density can vary with different layouts.

Lastly, forest conservation projects (especially for carbon offsets) are hindered by the efficiency, cost, and accuracy of tracking progress; our solution addresses these issues, thus catalyzing more conservation projects, leading to greater carbon sequestration in forests.

We have developed our v0 product for below canopy data collection (LiDAR+camera), and have collected our first datasets. We have begun developing our solution for above canopy data collection (drone + LiDAR), evaluating different sensor hardware for the v1 product, using an AWS backend for our data, and will develop our front end and AI algorithms over the next 3 months. We also have an LOI from Pachama, a Series A startup in the forestry space, so this development is geared towards meeting Pachama’s needs for higher resolution data.

To measure and monitor forests, foresters either use satellite data or manual data collection. High spatial resolution satellite images can have resolutions from between 50cm-5m (and only sample from above-canopy), while LiDAR can yield a resolution in the millimeter range (above- and below-canopy samples). To get more accurate data than a satellite, foresters generally use basic equipment like calipers, because the processing of a LiDAR point cloud requires robust algorithms that foresters do not have. The cost of sending out foresters to manually collect higher-resolution data on forest plots at scale is prohibitive for many players (could be upwards of $50k depending on number of plots, excluding most small land owners). 

The unique insight we bring is from the autonomous vehicle space, where Peter spent several years working: by taking a similar “algorithm stack” that an autonomous vehicle uses to segment, associate (among multiple sensor observations), classify, and estimate a state for objects in an urban environment, we can similarly process LiDAR data from a forested environment. So, the key innovation is the application of this end-to-end AI algorithm stack to the new domain of Forestry.  With this, we can provide biomass information with greater precision (mm vs. cm/m) than satellite data. We address the prohibitive costs by fixing a LiDAR to a drone for above-canopy measurements and automating full end-to-end data processing, enabling land owners to more easily monetize conservation of their forests through high-quality carbon offsets.

The impact our service will have will be on resolving a weakness in the carbon credit market.  We will help buyers have confidence that a carbon credit truly represents one ton of carbon by measuring the forestry assets with high accuracy using LiDAR and computer vision.  In helping the carbon credit market become more effective, and forests become more valuable assets, we believe our solution will help the whole of the Earth as nature is able to sequester more carbon to maintain the current climate equilibrium, and continue to support Earth’s current ecological systems.

Additionally, we hope to support locals in developing worlds by hiring locally for technical positions that require teachable skill sets to support our local deployments.

Our product is not yet in the market, but we have begun tracking indicators to validate both the market and the technology.

For market validation we have tracked two metrics: number of interviews, and LOIs. We’ve had almost 100 interviews with forestry stakeholders, many of whom have introduced us to potential customers. With these intros, we have narrowed our “first customer pipeline” to 5 firms. We are now moving forward with Pachama, who sent an LOI email for our data acquisition service for a pilot this summer.  The pilot with Pachama will bring us to measure 300 of the 1000 hectares we aim to measure in the first year. Moving ahead, we will additionally start tracking signed contracts, revenue, and NPS to measure product-market fit.

Towards validating the product, we have tracked signs of external validation of our approach, in addition to technical milestones. To this end, we have a partnership with a leading LiDAR company, and have won 2nd place in an MIT Deep Learning competition. Regarding technical milestones, we’ve developed our v0 product for below canopy data collection (LiDAR+camera), and have collected our first datasets.

Over the long-term, we will start tracking impact metrics as described above - including, but not limited to, area of forests we have measured/verified for carbon offset projects, and the amount of carbon sequestered in these forests.

5 people:

• 1 full time co-founder,
• 3 part time co-founders
• 1 part time advisor

Our team includes people with deep experience in robotics, from someone with five years of experience working on perception AI for autonomous vehicles and a masters degree in robotics, to another with a masters in computer vision and years of entrepreneurial experience both as CTO as well as working with data science, to one cofounder with a masters in Mechatronics and hardware-hacking experience as a board director at the biggest hackerspace in Detroit.  Additionally, one of us is at MIT Sloan pursuing an MBA and another pursuing an MBA at Harvard Business School, giving us a solid education around both business principles as well as skills towards creating an early startup.

We are attending workshops at Cleantech Open on how to improve diversity on the early team, and how to appropriately and effectively recruit for diversity.  We are determined to improve the diversity of our team, both for the longer term sustainability of our company in making a broader impact on the world, but also for the immediate, near term goal of benefiting from more diverse perspectives.

From Solve, we hope to engage with MIT mentors to help refine our go-to-market and scaling strategies. Since data is at our core, we specifically hope to work with the mentors to understand how we can use high resolution data to build a moat from potential competitors and to drive continued growth and value for our customers.

Second, since we are driven by our vision to make a meaningful impact towards fighting climate change, we hope to receive mentorship and tools to help measure our impact to hold ourselves accountable.

Lastly, with one LOI for a pilot in our hands, we hope to receive a small amount of capital to fuel these early pilots, which require drone and sensor hardware for execution. Even a small amount of capital will help us acquire the appropriate hardware, which we can use for a pilot with Pachama and any of our future pilots as well.

Beyond what we hope to receive, we also bring several things to the table. For example, we have access to several universities’ networks (CMU, Michigan, MIT, Harvard), and Peter brings a technical background in robotics, autonomous vehicles, and computer vision, and we could support entrepreneurs in any of these spaces with our experience and networks.

Human Capital: we are confident with our networks (from Harvard and Yale for the forestry expert, and from MIT, Carnegie Mellon and University of Michigan for the roboticist), but we need to give this serious effort and intentionality to find the perfect people for the team.  Rounding out our team is a top priority, since this work will bring compounding results over the next year as they contribute to the cause.

Business Model: We could use support to refine our go to market and scaling strategies. Since data is at our core, we specifically hope to understand how high resolution data can create the most value for forest managers.

Product/Service Distribution: While we have an LOI from Pachama, we are hoping to quickly scale our pilots to more early customers, and we could use support identifying and connecting with more potential customers.

We would like to partner with The Nature Conservancy, since they are at the center of this ecosystem and have many forestry projects they are directly involved in.  Their insights, as well as broad view of the forestry space, are uniquely valuable.  Additionally, we would like to get a large company such as Microsoft, Amazon, or Suzano onboard early on, as both a partner and potential customer, in order to benefit from their insights around offsets and forest management, as well as potentially work our way towards a scalable deployment.

The practice of measuring and verifying forest growth/conservation is both expensive and slow. In addition to all this, there is a lack of available data for project developers to intelligently design and manage forests. Furthermore, the lack of high resolution data on forests has instilled a lack of trust in forest carbon offset projects, limiting the potential of forest conservation as a way to tackle climate change. Our solution can create a new industry-wide best practice to use advanced technology to measure forests with greater accuracy and precision.

As our solution will lead to more forest conservation, communities around forests will also see a benefit in the health of their surrounding ecosystem, as well as greater potential for economic growth (as our responses note, we plan to engage local communities to teach them to use our solution to measure the forests, thus creating jobs within local communities).

With this prize, we will be able to acquire the hardware necessary to execute our pilot with Pachama as well as any future customers/pilots we acquire over the next year.

The practice of measuring and verifying forest growth/conservation is both expensive and slow. In addition to all this, there is a lack of available data for project developers to intelligently design and manage forests. Furthermore, the lack of high resolution data on forests has instilled a lack of trust in forest carbon offset projects, limiting the potential of forest conservation as a way to tackle climate change. Our solution can create a new industry-wide best practice to use advanced technology to measure forests with greater accuracy and precision.

With this prize, we will be able to acquire the hardware necessary to execute our pilot with Pachama as well as any future customers/pilots we acquire over the next year.

First, we are addressing the problem of efficient, accurate data collection in forests to measure biodiversity and carbon content. Our solution uses sensors and AI similar to autonomous vehicles to collect and process forest data into meaningful metrics. Furthermore, we use machine learning to classify trees and extract insights about forest layouts, including how resiliency, biodiversity, and carbon density can vary with different layouts.

With this prize, we will be able to acquire the hardware necessary to execute our pilot with Pachama as well as any future customers/pilots we acquire over the next year.