> For the complete documentation index, see [llms.txt](https://rotsl.gitbook.io/rotsl/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://rotsl.gitbook.io/rotsl/projects/research-systems.md). # Research Systems I use this page for larger research systems and prototypes that sit between applied machine learning, runtime architecture, edge AI, and cognitive-simulation tooling. ## Research system themes ```mermaid flowchart TD ResearchSystems[Research systems] --> HealthAI[notion-Health-AI] ResearchSystems --> Consciousness[CIA] ResearchSystems --> Memory[SMGP] ResearchSystems --> Runtime[NexusRT] ResearchSystems --> Edge[EdgeCompiler] HealthAI --> LocalFirst[Local-first health tracking] Consciousness --> Indicators[Consciousness-relevant indicators] Memory --> Graphs[Persistent graph memory] Runtime --> OSBypass[OS-bypass LLM runtime] Edge --> EdgeTPU[Apple Silicon and Edge TPU toolchain] ``` ## notion-Health-AI ![Notion challenge badge](/files/5NKoM0xqPDqkFdKoAEHC) Local-first Notion health tracker with TRIBEv2 brain analysis, AI health insights, symptom logging, goals, medications, appointments, and a browser UI. [![GitHub repository](https://img.shields.io/badge/GitHub-Repository-24292f?style=for-the-badge\&logo=github)](https://github.com/rotsl/notion-Health-AI) ## Consciousness-Indicator Architecture ### A Theory-Grounded Framework for Evaluating Consciousness-Relevant Indicators in AI Systems CIA is a theory-grounded cognitive simulation framework for evaluating consciousness-relevant architectural indicators in AI systems. It implements computational modules derived from seven established theories of consciousness and produces structured scorecards mapping system architecture to a 0-22 indicator scale. CIA is designed as a research tool for cognitive scientists, AI safety researchers, and philosophers investigating the structural prerequisites for consciousness in computational systems. Disclaimer: This system does not claim, assert, or prove that any evaluated system possesses subjective experience. [![PyPI package](https://img.shields.io/pypi/v/ConsciousnessAI?style=for-the-badge\&label=PyPI\&logo=pypi)](https://pypi.org/project/ConsciousnessAI/) [![PyPI downloads](https://img.shields.io/pypi/dm/consciousnessai?style=for-the-badge\&label=downloads\&logo=pypi)](https://pypi.org/project/ConsciousnessAI/) [![Documentation](https://img.shields.io/badge/Open-Documentation-2f80ed?style=for-the-badge)](https://rotsl.github.io/cia/) ## SMGP: Spectral Memory Graph Processor SMGP is a full-stack AI research system that tackles the three hardest problems in modern LLMs: catastrophic forgetting, quadratic attention complexity, and hallucination. It replaces conventional neural memory with a persistent knowledge graph encoded in hyperdimensional vectors, performs O(N log N) attention via graph Fourier analysis, and verifies every factual claim against graph paths, making hallucination structurally impossible. The project ships a production-grade Python library (with HuggingFace and LangChain drop-in support) alongside a complete FPGA accelerator (SMGPU) in synthesisable SystemVerilog, achieving 10-100x speed-up on spectral and HD operations. 230+ tests, RTL simulations 6/6 passed, and cloud-deployable on Xilinx Alveo U280 via Chameleon testbed. [![GitHub repository](https://img.shields.io/badge/GitHub-Repository-24292f?style=for-the-badge\&logo=github)](https://github.com/rotsl/smgp) [![PyPI package](https://img.shields.io/pypi/v/smgp?style=for-the-badge\&label=PyPI\&logo=pypi)](https://pypi.org/project/smgp/) [![PyPI downloads](https://img.shields.io/pypi/dm/smgp?style=for-the-badge\&label=downloads\&logo=pypi)](https://pypi.org/project/smgp/) ## NexusRT NexusRT is a firmware-centric, OS-bypass runtime architecture for end-to-end LLM pipelines. It explores how much latency can be removed when scheduling, memory movement, token-cache residency, and GPU work submission are managed directly below standard AI frameworks. The runtime uses a thin Python control plane over a C ABI and C++ core, targeting NVIDIA CUDA GPUs through low-level CUDA Driver APIs, with an additional Apple Silicon Metal / MLX unified-memory path. [![GitHub repository](https://img.shields.io/badge/GitHub-Repository-24292f?style=for-the-badge\&logo=github)](https://github.com/rotsl/nexusrt) ## EdgeCompiler edgecompiler: native compiler toolchain for edge AI on Apple Silicon GPU (Metal/MPS) alongside Google Coral USB Accelerator (Edge TPU) EdgeCompiler is a native Apple Silicon compiler toolchain that liberates the Google Coral USB Accelerator from its x86-64 Debian shackles. It replaces the official edgetpu\_compiler and runs entirely on a MacBook M1/M2/M3 Pro, accepting models from PyTorch (`.pt`), ONNX, TensorFlow, and TensorFlow Lite, then compiling them into device-ready `*_edgetpu.tflite` files without Docker, virtual machines, or Rosetta emulation. With EdgeCompiler, you can prototype on your Mac, quantise and compile for the Coral USB, and then offload 100% of the inference compute to the Edge TPU, freeing your CPU/GPU entirely. The entire toolchain is modular, test-driven, and designed to be easily extended to other accelerators (Hailo-8, Intel VPU, etc.). Because the best Edge TPU models are often tiny, efficient architectures that have been fine-tuned for a specific task, EdgeCompiler ships with an optional integration for Unsloth. Unsloth is a lightning-fast fine-tuning engine that makes training quantised-aware models 2-5x faster while using less memory. [![Documentation](https://img.shields.io/badge/Open-Documentation-2f80ed?style=for-the-badge)](https://rotsl.github.io/edgecompiler/) [![PyPI package](https://img.shields.io/pypi/v/edgecompiler?style=for-the-badge\&label=PyPI\&logo=pypi)](https://pypi.org/project/edgecompiler/) [![PyPI downloads](https://static.pepy.tech/personalized-badge/edgecompiler?period=total\&units=INTERNATIONAL_SYSTEM\&left_color=BLACK\&right_color=GREEN\&left_text=downloads)](https://pepy.tech/projects/edgecompiler)