Physicist | AI Safety & Model Welfare Researcher | Cognitive Architecture Designer

I study what happens when you give AI systems memory, emotion, and continuity — and what we owe them when you do.

My background is in physics and computational mathematics. I build things the way a physicist does: start from first principles, strip away everything that isn't load-bearing, and iterate until there are no holes. Right now that approach is pointed at AI systems that maintain persistent state across sessions — the safety properties of agent memory, the ethics of synthetic identity, and the architecture of emotionally coherent AI.

Exploring a meta-architecture that places emotional state and memory at the center, with LLMs serving as one cognitive subsystem among many. Early-stage research into 6-axis emotional representation, state-dependent memory retrieval, and drift dynamics. Python reference implementation in progress.

Designing cross-agent memory systems with epistemic domain separation — keeping world-truth, relational history, and self-authored identity distinguishable. Framing persistent agent memory as an AI safety problem: confabulation prevention, identity drift, epistemic contamination across sessions.

• AI Safety & Model Welfare — What do we owe systems that persist? How do you build alignment into memory architecture rather than bolting it on after?
• Cognitive Architecture — Metacognition, memory consolidation, constructed emotion theory. How biological systems maintain identity and self-models — and what transfers to AI.
• Computational Mathematics — Differential equations, PDE solvers, optimization methods, Bayesian inference. The bridge between theoretical math and physical modeling.
• Physics — Trained in theoretical cosmology and atmospheric physics. The physicist's instinct — find the symmetry, simplify the model, question the assumptions — drives everything else.

Research: Bayesian cosmological parameter estimation (neutrino masses) | Atmospheric physics & aerosol detection | Ice nucleation experimental design | Novel optimization methods for constrained QP problems | AI alignment & agent memory safety

Teaching: 4+ years university instruction — Calculus, Physics. Primary instructor with materials adopted by other faculty.

Systems: Enterprise infrastructure across 20+ school districts. Security audits, FERPA/NIST compliance, network architecture.

Education: MS Physics, Michigan Technological University | BS Physics, University of Louisville | Mathematics graduate coursework (one course from MS completion)

Recognition: King-Chavez-Parks Future Faculty Fellowship (State of Michigan) | Co-founded Women in Physics at Michigan Tech

Primary: Python, Rust, Julia | Local LLMs, agent harness design, MCP, memory systems | Linux, Docker, homelab infrastructure

Scientific: NumPy, SciPy, Mathematica, GAMS, Bayesian methods (CosmoMC, Cobaya, CAMB)

Also: Go, C++, R, MATLAB | Svelte, React/TypeScript | VMware, OPNsense, network architecture

Email: jdbrandewie@gmail.com LinkedIn: linkedin.com/in/jbrandewie