Accelerating conservation research with AI
For over six decades, the team at the Jane Goodall Institute’s Gombe Stream Research Center in Tanzania has collected daily data records on a checkbox-style document called a Tiki sheet. The researchers follow a single, “focal” chimpanzee all day and record data points that include the arrival and departure times of other chimpanzees into the focal chimpanzee’s subgroup, feeding bouts and encounters with other species.
Over the years, hundreds of thousands of records have been collected and are stored in the Jane Goodall Institute's Gombe Research Archive on ASU's Tempe campus.
A team of student researchers has been manually entering the information collected in the sheets into the project’s database. In fall 2025, Gilby turned to the university’s AI Acceleration team for support.
Senior AI development engineer Krishna Sriharsha Gundu, part of the AI Acceleration team at ASU, took on the challenge.
He worked toward a solution that would combine machine learning with imaging techniques called “vision” to review the tracking sheet and analyze the data on the page. The team coined the project “Gombe AI.”
Taking a scanned image of the Tiki sheet, the computer vision software straightens the image and extracts the data points. The software is then able to digitize the image into simple rows and columns on Excel spreadsheets. These data will be incorporated into the project’s relational database for analysis.
“We combined this traditional AI technology with newer large language models to review and analyze the handwritten notes written in the margins on the sheets,” Gundu said.
Gundu developed the code needed to extract data from the Tiki sheets, data which student researcher and data science undergraduate Joesh Jhaj then takes and focuses on the project’s interpretation and translation.
“Computer vision translates an image that the computer can then pick up on,” Jhaj said. “The use of generative AI in this project is to read and translate the handwriting.”
Building on Gundu’s computer-vision preprocessing, Jhaj wrote the translation and interpretation code that turns contextual marks into structured research data, using GPT’s API to verify unclear handwriting or ambiguous symbols when needed.
After the data is digitized, students who work with the Gombe chimpanzee archive will check and compare the results against the original handwritten data sheets.
Jhaj is in the early stages of working with the new Gombe AI solution. The team hopes the tool will improve accuracy in its analyses and be less time-consuming for data entry. The tool will also enable researchers to more effectively connect, manage and update the Tiki sheet data with the other handwritten protocols, video and geospatial data digitized as part of the Gombe AI Research Platform.
“It’s awesome to be able to work with teams across the university and develop unique AI solutions that advance research like this,” Gundu said.