Soju06

Codex/ChatGPT multiple account load balancer & proxy with usage tracking, dashboard, and OpenCode-compatible endpoints

Layered, replicated, self-adjudicating code review orchestrator

Docker images for machine learning development environments using CUDA and PyTorch and for remote development via VSCode and SSH server

1,100+ FPS YOLOv8/YOLO11 inference on a $400 GPU — every kernel hand-written in one CUDA file. No cuDNN, no TensorRT, no Python at runtime. All scales, all task heads, same detections.

LLM knowledge cutoff compensator for AI coding agents — corrects outdated versions, deprecated APIs, renamed packages, and removed CLI commands

Hand-rolled BGE-M3 inference engine: multilingual embeddings at 1800 req/s on a $400 GPU. Zero-dependency C + CUDA, beats PyTorch sdpa at every length.