about

AI/ML engineer and systems programmer working across the full stack — from PCIe/DMA hardware interfaces to transformer inference and diffusion pipelines. VFX background. I build things that think, then make them run fast on real hardware.

focus

🧠 LLM Inference

vLLM · llama.cpp · CUDA kernels · KV-cache optimization · speculative decoding

🎨 Diffusion Models

SDXL · ComfyUI pipelines · LoRA training · VFX integration

🔧 Systems

DMA/PCIe · memory-mapped I/O · kernel modules · WireGuard mesh

🤖 Multi-Agent

AutoGen · LangChain · tool-calling · structured output · edge deployment

stack

Python C++ Rust CUDA PyTorch Transformers vLLM TensorRT ONNX ComfyUI Docker Linux

currently

neurologist-7b-instruct — Fine-tuned Llama for technical reasoning and code generation
sdxl-vfx-cinematic-lora — LoRA trained on VFX reference frames for cinematic generation
Edge inference — Sub-200ms serving on consumer hardware (9950X + RTX)
Homelab mesh — 4-node WireGuard cluster for distributed inference and automation

Neurologist