Revital Kremer Optimizes Crop Yields for Farmers by Harnessing GenAI at SupPlant
Revital Kremer, CTO at SupPlant
CTO at SupPlant
Revital Kremer is leading the charge for ag-tech to help farmers overcome water scarcity as climate change continues to challenge growing conditions around the world. She joined SupPlant, a precision ag-tech company, after more than two decades working with technology teams in military, gaming, and other software domains.
SupPlant’s system gathers data from sensors placed on the plant trunk, on the fruit, and in the soil, collecting and transmitting measurements in real time. Learn how Revital and her team help farmers grow healthier plants through AI-powered irrigation, backed by real-time data.
Transcript
At SupPlant we like to say that we help farmers speak better plant. This is what we are good at, translating data from plants into actionable insights and recommendations and explain the farmers what happened, what exactly happened with the plants in real time, everything begins in the field where our sensors are located at.
We have an IoT solution that collects data, transmit data to the cloud every half an hour. This data is manipulated and observed right away from the IoT solution and reflected in our application with the recommendation, with the data and so on. This is one solution that helps farmers understand what's happening in the field in near real time. So this is one product line that we have. It's called Sense. The other product line is called Plant. It's focused on smallholder farmers that do not have the ability to pay for the cutting edge solution and it's based on the sensors, the big data, but works as sensorless solution.
GenAI is the new kid on the block where we can utilize large language models for benefits in order to combine our big data with the professional knowledge to create what we call the AI doctor or the professional agronomist to embedded into the system.
We get used to work with Astra DB and vector search is a permutation of Astra DB. It's very intuitive to choose this technology and on the other hand it was very easy to create the embedded data using vector search.
We split the POC into two phases once we have the problem or the question how to query the vector search DB for the right documentation to be injected into the LLM model. This is the first phase.
The second phase is to analyze what is the problem which we are in very early stages of this phase. We tend to take a screenshot of the visible screen that the user see and analyze it with multiple models to understand what happening currently with the data that the users see.
I really like the fact that for every person that I talk with, I used to talk with in DataStax the conversation ends with okay, so what more can I do for you? Which is amazing. It's a great experience. So it's all about people and good technology.