Lead researcher profile
I led AlphaHand from the first idea through implementation. My role spans research design, EEG and machine-learning development, participant coordination, data analysis, robotic-hand assembly, public reporting, personal funding, and outreach.
Figure 1: Signing a compute module at the DENSO Maryville manufacturing plant after meeting with autonomous-driving researchers. (Left to right: Sen. Lamar Alexander, Jonathan Davanzo, and Matthew Davanzo.)
As lead researcher and developer, I connected the project across research, engineering, hardware, communication, and participant logistics. Beyond building a demonstration, I built the evidence trail that lets others inspect the data, metrics, reports, model-selection history, and limitations.
A friend born without a hand, a family I met at ISEF who was navigating Alzheimer's, and accounts from around the world about how physical and neurological challenges affect everyday life. Those experiences have pushed my work toward outreach, education, and practical impact.
Public scale
These numbers come from the public AlphaHand result bundles and update trail. They describe the documented project evidence, not every private experiment attempted while building the system.
2 public subject bundles
The site currently publishes 1-M16 and 2-M16 bundles with metrics, figures, and HTML reports.
4,953 test windows
The published test splits include 2,652 windows for 1-M16 and 2,301 windows for 2-M16.
12,447 replay windows
The featured 2-M16 checkpoint is also checked on a cleaned pseudo-live replay corpus.
2,595 configs
A documented deployment-facing ablation compared thresholds, smoothing, hysteresis, adjacency, and finger-mode settings.
100+ trained variants
The project history documents a 30+ hour tuning sweep, including a 90-run block logged over 33.3 hours.
95.37% precision
Among non-rest replay windows that passed the current actuation gate, 95.37% carried the correct movement command.
Contributions
I defined the original assistive-technology goal from personal experiences, scoped the EEG action-state and active-finger decoding challenge, and planned how data, models, and physical actuation would connect.
I researched the signal-processing and machine-learning approach, then wrote the training, evaluation, replay, and reporting code that makes the project auditable.
I coordinated human participation and data-collection sessions, keeping the resulting EEG evidence organized around public run bundles, metrics, reports, and figures.
I analyzed the collected EEG data, compared model behavior across documented tuning and ablation campaigns, and translated the results into public metrics and explanations.
I assembled and integrated the robotic hand so decoded EEG outputs could connect to a physical assistive-device demonstration.
I personally funded the entire project using earnings from my summer job as a STEM-focused camp counselor.
The research process changed the project from a technical build into a way to support and explain science in the community. Meeting people affected by neurological and physical challenges made the work feel less abstract, and my outreach grew from showing how AlphaHand works to helping more people see research as a practical response to real human problems.