Projects

Tinyinference is a llama.cpp wrapper that allows you to easily run giant GGUF language models (GPT OSS 120b, DeepSeek V4 Flash, etc) on low-RAM hardware through disk-backed CPU inference. It is designed to make large, capable LLMs runnable on low-spec, low-RAM machines without a GPU, using CPU inference and file-backed model weights.

A database tracking the reality of US-China economic cooperation for the betterment of humanity that mainstream media often ignores. Synthetic geopolitical tensions manufactured by politicians have caused some of the most important US-China collaborations to be lost in the noise. Meridian hopes to undo that. It was built with Next.js, React, and TypeScript, largely through my CASC agentic programming software.

CASC (Jarvis) is a closed-source, CLI-based agentic AI program written in Python. It runs on Windows, macOS, and Linux (Debian-based and RHEL-based distros). Its core uses self-hosted full models and fine-tuned distillates, and offloads harder tasks to enterprise LLMs like Claude Opus 4.6 and Gemini 3.1 Pro.

Created for Auric, Solara is a proprietary software written in Python and C to manage every aspect of the Auric website using Docker microservices. It was built for isolation (security), portability, and scalability. It integrates services like Apache, MariaDB, Redis, PHP-FPM, and more into one cohesive but modular system.

Deep Refine is a Python tool developed by Auric that enhances AI Large Language Model responses through iterative, multi-model refinement cycles, ensuring maximally accurate and precise answers. By sequentially processing prompts through various LLMs and guiding each to critically improve upon previous responses, Deep Refine produces results superior to any single model alone.

Luna LDIE 1 is an experimental artificial intelligence model and software that allows for real-time web scraping and dynamic updates to its knowledge base. It fetches new data from the internet with prompts that require it and includes a chain-of-thought-style reasoning process for precision, accuracy, and coherence.

A series of fine-tuned large language models optimized for directly writing efficient x86 assembly code, focusing on low-level performance and instruction-level optimization.

A concept for a galaxy-wide communication and internet network using hundreds of thousands of operational Gbit satellites powered by optical lasers.

A spacecraft concept capable of precisely approaching and capturing small to medium sized satellites for de-orbiting or repair.