Highlights

NASA L'SPACE systems engineering, USC research labs, and independent research projects—from the curriculum vitae.

Experience and research projects from my curriculum vitae—NASA L'SPACE MCA and NPWEE, plus independent builds in agentic AI and medical ML.

NASA & Aerospace

Research

Engineering

Projects

NASA & Aerospace

NASA

L'SPACE MCA — Team 19: Ad Astra

2026

Lead Systems Engineer on an 18-member MCA team designing an autonomous lunar PSR volatile-prospecting rover at Nobile Rim 2.

From the journal: MCA Team 19: Ad Astra formed around a concrete science objective: in-situ volatile characterization in a lunar PSR. I stepped in as sole Lead Systems Engineer.

NASA

L'SPACE NPWEE — Team 10: Forged in Orbit

2025

Lead Systems Engineer on a 12-member NPWEE team—every engineer answered to me—authoring a $10,000-class cold-welding proposal and lead-authoring L-MAP-FI.

From the journal: NPWEE Team 10 targeted solid-state in-situ metallurgical reconstitution via autonomous cold welding. As Lead Systems Engineer, every engineer on the team answered to me—including Engineer 1 Justin Schueler—while I coordinated with PM Alexis Gallardo and Chief Scientist Riya Jain on programmatic and science integration.

Projects

Personal project

SpaAIder — Agentic Runtime

2025–Present

Local-first multi-agent runtime with persistent cross-session memory, heterogeneous clients, and Ollama-first sovereign inference.

From the journal: SpaAIder deliberately splits concerns: the TypeScript control plane runs runOrchestrator; the Python FastAPI substrate owns SQL tables, episodic store, and /api/v1/memory/recall.

Personal project

Pneumonia Detection from Chest X-Rays

2025–2026

TensorFlow pipeline with EfficientNetV2B0 transfer learning—83% validation accuracy on 368 held-out radiographs.

From the journal: The CV documents 988 chest X-rays split 620/368 train/validation. I built train.py around EfficientNetV2B0, not a from-scratch CNN—data volume does not support training millions of weights cold.

Personal project

LLM-V1.0 — Small-Scale Language Model

2025–2026

Designed and trained a ~30M-parameter GPT-style transformer (8 layers, 8 heads, 512-dim embeddings, 8,192-token BPE vocab).

From the journal: At ~30M parameters, LLM-V1.0 is intentionally trainable on student hardware. Eight transformer layers with eight heads and 512-dimensional embeddings balance capacity vs. overfit risk on modest corpora.