southerncrossai/docker repository overview

⁠🌌 Southern Cross AI | JoeyLLM Official Docker Image

Welcome to the official Docker repository for JoeyLLM, engineered by the research team at Southern Cross AI.

Based out of Australia, our team is dedicated to pushing the boundaries of foundational models with a focus on high-efficiency training, bare-metal hardware optimization, and seamless scalability from local testing to massive Kubernetes clusters.

This container provides a fully configured, ready-to-deploy environment for running, fine-tuning, and inferencing JoeyLLM, completely pre-packaged with all necessary CUDA drivers, PyTorch dependencies, and optimized inference engines.

---

⁠🦘 What is JoeyLLM?

JoeyLLM is a next-generation Large Language Model architected from the ground up to punch above its weight class. Designed by Australian researchers for global applications, JoeyLLM focuses on:

• Hyper-Optimized Throughput: Engineered to squeeze every drop of compute out of your GPUs, minimizing idle time and maximizing token generation.
• Scalable Architecture: Whether you are testing locally or deploying across multi-node enterprise environments, the model dynamically scales to fit your hardware.
• Robust Alignment: Built with safety, accuracy, and advanced reasoning capabilities at its core.

---

⁠🛠️ Built for Production

At Southern Cross AI, we know that building the model is only half the battle. This container is built with production workloads in mind:

• Kubernetes Ready: Includes health check endpoints (/healthz) and graceful shutdown handling for flawless pod orchestration.
• Observability: Exposes native Prometheus metrics on port 9090 so you can monitor token latency, batch sizes, and GPU thermal throttling in real time.
• Immutable Infrastructure: Statically linked binaries ensure that what you test in staging is exactly what runs in production.

---

⁠🌐 Connect with the Team

We are always looking to collaborate with other researchers and engineers.