Peter Seelman

I am a human studying Physics and Mathematics at Johns Hopkins University. I am currently building Altera Labs.

My focus is on AI agent architectures and knowledge systems for education and national security. Below is a selection of my recent work.

Selected Work

Altera Labs — A state-aware AI tutoring platform for STEM, using knowledge graphs and Bayesian tracing to teach rather than give answers.
APL — AI system architect at Johns Hopkins Applied Physics Laboratory, building agent orchestration and evaluation infrastructure for national security workflows.
UNITE — Founded a $35K-funded program formalizing the JHU-APL talent pipeline through an Intersession course with lab tours, executive speakers, and recruitment pathways.

Reach out via email or find me on GitHub and LinkedIn.

Things I keep coming back to. Not finished thoughts — just directions.

The substitution problem — Students use AI to skip thinking. The tool that was supposed to help is making them worse. How do you build something that resists this?
Knowledge as a graph — Understanding isn't a score. It's a structure — prerequisites, dependencies, gaps. Can you map it and use it?

Agent orchestration — Multi-step AI that does real work. The hard part isn't the model — it's knowing when it's wrong.
Building in public — What it's like to start something from scratch, make it real, and figure out if anyone cares.

I wrote a paper on how AI interfaces with human cognition — it became the research behind Altera.

Surely You're Joking, Mr. Feynman!

Half brilliant thoughts on physics and curiosity, half questionable thoughts about women. The first half stuck with me.