Available for work
I research machine learning, build high-performance infrastructure, and write about where the technology is heading.
01About
I'm a high school student in Cleveland, Ohio who got into machine learning and software the way most people fall into a hobby that quietly takes over.
Most of what I know came from building things, breaking them, and figuring out why. Lately that means interpretability research on audio foundation models, low-level performance work, and the occasional essay when an idea won't leave me alone. I like problems that are technically hard and actually useful.
02Experience
With a research team · Remote
Interpretability and structured sparsity for audio foundation models. I probe which internal representations actually carry the signal, then prune aggressively to see what survives. Writeup in progress and under review.
StackPyTorch · activation probing · structured + magnitude pruning
Cleveland Clinic · On-site
Working with the IT infrastructure team on the systems that keep a hospital network running, across software infrastructure and enterprise information technology.
FocusEnterprise IT · network infrastructure · hospital systems
Solon Senior Center · Part-time
Supported 50-plus seniors across phones, laptops, and car infotainment systems, ran VR sessions on the Oculus Quest 2, and built training modules that measurably improved digital confidence over six months.
ToolsOculus Quest 2 · device support · training modules
Mitchell's Ice Cream
03Projects
Research
Finding the representations that matter inside audio foundation models, pruning the rest, and measuring how much you can take away before the model notices. Under review.
StackPyTorch · probing · activation patching · structured sparsity
1st Place · Scrapyard Cleveland
A machine with exactly one purpose: turning itself off. Tap the RFID tag, the button arms, and a servo arm flips it back off in about 50 milliseconds. Firmware in C++ on a microcontroller. It does nothing, precisely.
BuildC++ firmware · RFID · servo arm · ~50 ms off-time · built in 24h
04Writing
On how schools respond to AI · 2025
A 70% flag from a probabilistic classifier is a hypothesis, not a finding.
An argument that the most important effect of generative AI on education isn't the technology — it's how institutions react to it. Detector false positives, Bloom's 2-sigma problem, and the case for integration over prohibition.
Read the essay ↗Aside —Bloom's 2-sigma problem (1984): one-on-one tutored students outperform 98% of a conventional classroom.
Rising Scholars Award · 2026
If a million-qubit computer can be built, why not 10 or 100 million?
What changes when qubits stop making mistakes. From Majorana 1 to PsiQuantum's roadmap, a look at the race toward a million fault-free qubits and what boundless quantum compute would mean for medicine, cryptography, and physics.
Read the essay ↗Aside —Majorana 1 is Microsoft's topological qubit chip; PsiQuantum is pursuing a photonic path to the same scale.
05Education
Senior · Class of 2027
06Skills
07Contact
Open to research, engineering, and a good problem.