Ishaan Dokania, a 12-year-old sixth grader at Willamette Valley Academy, is competing in Washington, D.C., as one of 30 national finalists in the Thermo Fisher Scientific Junior Innovators Challenge.

Dokania built a machine learning model that uses satellite imagery to pinpoint lithium deposits, achieving 89 percent accuracy. Lithium is a critical mineral for batteries and energy storage, but locating new sources is expensive and can damage the environment, according to the Society for Science, which runs the competition.

The Beaverton student's interest started during a Science Olympiad competition, where he took on two topics: remote sensing and rocks and minerals. He wondered whether he could combine the two to solve a real-world problem and learned about the global lithium bottleneck.

To build his model, Dokania trained it on satellite image data from Google Earth Engine and mineral data from the United States Geological Survey Mineral Resources Data System. He tested several algorithms and identified sources of noise that threw off his results. His models produced biased readings when they detected plants under heat stress, flagging any hot region as a lithium site. He added a test called Leave One Feature Out to filter that noise.

He also found that using different satellites for data could produce overlapping maps, another accuracy problem he had to solve.

The result: 89 percent accuracy in detecting lithium deposits.

Dokania presented his research at a public showcase at Planet Word Museum in Washington on Saturday, Oct. 24. He told the Society for Science that the experience has connected him with peers working on similar problems.

"For example, there's another finalist here and we both used the same machine learning algorithm in our research, but his model is trained to predict on something else and now we're talking about different types of machine learning," Dokania said.

Outside the lab, he plays badminton with his dad and younger brother, keeping a running rally record. He wants to be an aeronautical engineer.

Patch highlighted Dokania's selection in its Oct. 25 Beaverton newsletter. Winners of the 2026 Thermo Fisher Scientific Junior Innovators Challenge are scheduled to be announced Tuesday evening, Oct. 27.

Upcoming in Beaverton: Free Electronics Recycling Event, 9700 SW Harvest Ct., Thursday, Nov. 5.