coreweave

Multi-cloud Terraform deployment framework for the vLLM Production Stack — with implementations for AWS (EKS), Azure (AKS), GCP (GKE), Nebius, and Oracle OKE.

Production-ready Kubernetes project for GPU-backed AI inference (vLLM, OpenAI-compatible API) on CoreWeave Kubernetes Service — Kustomize overlays, RWX storage, GPU node scheduling, NGINX ingress with TLS.

Cross a real highway as a chicken. Every collision is a real YOLO detection in VastDB; when a truck gets you, the agent hunts it across every camera and files the incident. Autopilot policy trained with RL and tracked in Weights & Biases. Built on VAST AI OS + NVIDIA Cosmos for the VAST Builders Challenge 2026.

VAST Builders Challenge: Scribner warehouse safety andon board detects near-miss hazards using Cosmos captions, YOLO11 detections, Embed1 semantic search on CoreWeave GPUs. Kubernetes deployment with live video, operator review, W&B retraining. https://luma.com/vastsf

SIEGE: 200 people vs one agent. A typed action gate (TypeSafe System One) that learns from every breach, evaluated by W&B Weave, hardened by a defender loop. Built at CoreWeave Hacks: Agent Loops 2026.

Runtime firewall for autonomous AI agents: allow, block, or require a human before any tool call touches infrastructure. Scar memory, governed self-improvement, CoreWeave sandboxes, W&B Weave evals.

PyTorch + Weights & Biases demo project for CoreWeave AI Solutions Engineer role

The AI Private-to-Public Ledger: pre-registered benchmark for venture marks and liquidity realisation. DOI 10.5281/zenodo.20672581

Ansible Collection for CoreWeave GPU cloud