SERES · ML Engineering · June–August 2026
Tactile sensing AI
& automated testing.
I led tactile-sensing AI development, designed the test bed, and developed the testing and calibration methods. The work connects controlled experiments and sensor characterization to a two-headed neural network that predicts contact position and force.
- Full-scale error
- <1%
- Model fit (R²)
- ≈0.99
Reported holdout results · research setup
Designed and automated the test bed.
I designed the test bed and wrote force-control software for a Torbal setup that could not regulate force on its own. Engineers could set a target load and run automated tests with built-in graphing and analysis.
Developed testing and calibration methods.
I researched and developed methods for choosing how many force reference points and contact locations to use, and for determining spatial resolution. I also defined the test matrix, test procedures, and calibration methodology.
Led development of the tactile AI.
I led the sensor’s AI development and improved the previous model. The two-headed neural network estimates contact position and force from magnetic sensor readings.
Validated position and force predictions.
I evaluated position and force predictions on a separate holdout run and documented the testing software, calibration procedures, and analysis workflow for continued data collection.