Alexander J Nagy

Technical Profile

I am a mechanical engineer and systems developer focused on high-performance computing and structural manifold analytics. My work at SEP Dynamics LLC centers on the Structural Manifold Engine: a text-only compressor/verifier that outperforms DeepSeek-OCR on Fox/OmniDoc benchmarks, and its first utility—FX regime tracking.

Background: production automation (Apple/Flex), rapid prototyping (CrunchLabs), and industrial product engineering. Provisional patent filed on the manifold approach; research artifacts are public and reproducible.

For a detailed history of production systems and engineering roles, visit the Experience page.

Technical Stack

Core: C++17/20, Python, CUDA, CMake, pybind11
Systems: Low-latency architecture, lock-free IPC, memory-mapped I/O
Research: Structural manifolds, information theory, FX regime analytics

Projects Portfolio

Structural Manifold Engine

Text-only structural compressor and verifier. 41–42× byte / 85–90× token compression on Fox/OmniDoc; 94.9–95.3% token accuracy; <0.09% FPR at 100% recall; completes in <1 hour on a single 3080 Ti.

Brief: Compression report

Riemann Pattern Evolution

Mathematical exploration of pattern stability (research prototype).
Repo: riemann-pattern-evolution (private)

Charge Routing Networks

Interactive visualizations of spiral network electrical properties, exploring resistance scaling and current distribution optimization.
Status: Research prototype; available on request.

FX Regime Tracking Demo

288-week ROC study on seven FX pairs. Mean-revert strands retain +28.9 bp at 360m; lead/lag signals lift neutral prediction to 55.7% (5y) / 69% (26w, high variance).
Brief: Signal-Regime whitepaper

See all projects »

Note: Private artifacts available on request under NDA.