ByMyCell, a Brazilian startup that applies modern genomics research to agriculture, is using Oracle Cloud Infrastructure (OCI) to help farmers and their suppliers maximize crop production sustainably.
Leo Leung, Vice President of Oracle Tech and OCI talks with ByMyCell founder and CTO Rafael Silva-Rocha about how technology can help. The interview was edited for length and clarity.
Leo Leung: What’s the startup environment like there?
Rafael Silva-Rocha: We are in São Paulo state, which accounts for about 40% of the Brazilian economy. A number of unicorns were born here. But one difference here is that if a Brazilian company wants venture capital, it must generate revenue from the beginning. It can’t say, “Oh, I have this nice technology, but no clients.”
Leo Leung: What problem is ByMyCell taking on? Who are your customers?
Rafael Silva-Rocha: We are a biotech company focused on simplifying genomics for agribusiness. My co-founder and I come from academic backgrounds dedicated to teaching and research. We started ByMyCell as we saw a way to build a technology solution to help Brazilian farms produce healthier crops. We target two kinds of customers – farmers and their suppliers. Farmers face challenges with inclement weather and in replacing chemicals with biologicals. Additionally, depending on the soil, certain combinations of soil and biologicals may not result in the expected crop yield. We provide a precision agriculture solution based on genomics microbiome analysis, which lists the microorganisms in the soil. With that information, farmers can make decisions based on data, not experimentation.
Second, our reports enable suppliers to tailor products to specific soil types. They can build products based on what is most appropriate for the soil, which helps the farmers grow healthier crops that yield higher growth. It is also good for the long-term health of the land.
By running genomics through our platform, they can select new microbes and deliver a new product up to six times faster than before. Also, our tests also show which biologicals will be most effective given each farm’s conditions.
Leo Leung: Before diving in, tell me more about the Brazil market?
Rafael Silva-Rocha: Agribusiness is one of the largest industries in Brazil, currently representing more than 20% of GDP and one-third of employment. Farmers are trying to reduce the amount of chemicals they use as they lead to soil degradation and impact plant growth. Switching from chemicals to biologicals—derived from naturally occurring materials—has the opposite effect, because it enhances soil fertility and promotes healthy development for a more sustainable and long-term solution, while also decreasing their carbon footprint. Our tests can show which biologicals will be more effective for their conditions. However, they can be trickier because it isn’t a broad-based approach and requires a farm to know the biological condition beforehand to correctly map to what products they use.
Leo Leung: What challenges do you see in terms of infrastructure or cloud services?
Rafael Silva-Rocha: The unfortunate side of working with AI and GPUs is that there are sometimes shortages—companies often compete for cloud resources. With Oracle it only takes 24 hours to release orders for new computing resources while with other cloud providers it can take a week. This speed is essential for delivering projects at the speed demanded by the customer. Additionally, Oracle offers superior price performance compared to other cloud vendors.
We decided to mix local and cloud. We generate the data and do the processing immediately; it reduces complexity and puts a lot of GPU computation there. And then we continue in the cloud. Having a hybrid infrastructure is okay for this stage, but more difficult to manage.
Leo Leung: I believe that you are using AI Infrastructure and HPC Compute. Can you share more?
Rafael Silva-Rocha: In the lab, the type of genomic sequence data in the machine gets data from an electrical signal, not the classical ATGC sequence. We convert the noisy electrical signal into a DNA sequence into text. To do that, we have models trained by specialized companies, and we must run them on GPUs. I may get 30GB of data for a small farm, and I must compress that 10x coming from this very noisy signal into a DNA sequence like text. We then upload that to the cloud and run the pipeline to convert the mix of sequencing into structured data, where we can see which kind of microbes we have and the relative abundance.
The first part is on GPUs in the lab right now, and then the final turn runs in the cloud on an Oracle HeatWave database. That second part, which is also high-performance computing, is done with CPUs on the cloud.
Leo Leung: What other OCI products are you using?
Rafael Silva-Rocha: We are starting to play with the Oracle HeatWave database and should expand usage shortly. Now, we must integrate this data in a more accessible and scalable way. And then, we must train new models to make predictions because, at the end of the day, our clients don’t want a report, they want a decision: “Okay, this is the picture of my land. What should I do? And if I do that, how much more can I gain in productivity?”
Leo Leung: Did you develop applications on OCI?
Rafael Silva-Rocha: It’s entirely OCI. We have one app that is an automatic proposal generator that lets users calculate samples. Another is a lab information management app that provides the infrastructure to track and process samples.
Learn more about the ISV AI service offerings on OCI to help innovators accelerate the development and deployment of production-ready AI here.
For more information on OCI’s AI services, GenAI, and cloud applications—all on a best-in-class AI infrastructure, click here.
