GPU Computing. ML Notebooks. One IDE.
A desktop IDE with a multi-cloud GPU marketplace, integrated JupyterHub, HPC cluster management, and a full terminal computer. Stop switching between five tools to run one experiment.
Compare GPU Pricing Across 9 Cloud Providers
Provision GPU compute from Hyperbolic, Google Cloud, Azure, Lambda Labs, CoreWeave, and more. See per-GPU pricing, configure clusters, and scale — all from one interface.
One Dashboard. Nine Providers.
Per-GPU pricing from Hyperbolic, Google Cloud, Thunder Compute, Azure, Hyperstack, Lambda Labs, CoreWeave, DataCrunch, and AWS. Select a provider, configure your cluster, and launch — without leaving the IDE.
HPC Cluster Orchestration
Manage Slurm clusters, submit jobs, monitor workloads. Connect AWS ParallelCluster, on-premises clusters, or your local GPU — one control plane for all compute.
Cluster Dashboard
Connected Clusters, Running Jobs, Pending Jobs — live metrics from your HPC infrastructure. Local Compute connection status, cluster management, job queue with per-job status tracking.
Local GPU in Three Commands
Install nexus-cli, start the compute agent, auto-detect hardware. Configure memory allocation limits and idle timeout preferences. Your local GPU joins the compute fabric.
JupyterHub with GPU Acceleration
GPU-accelerated ML notebooks integrated with the Nexus terminal, file browser, and HPC compute fabric. Persistent sessions, real-time output, full terminal alongside your notebooks.
GPU Notebooks + Terminal Side-by-Side
JupyterHub running GPU-accelerated notebooks alongside the Nexus Terminal Computer. GitHub file browser on the left, GPU compute output in the terminal, notebook in the center.
Research-Grade GPU Compute
Hyperbolic Dedicated GPU, DeepSeek-R1 via vLLM, MAPO Tournament optimization with specialized experts, MAP-Elites Quality-Diversity Archive — production ML research workflows in JupyterHub.
CVAT Integration for PCB Inspection
GPU-accelerated computer vision annotation with CVAT. Upload PCB images, define defect categories, auto-annotate components, and train ML models — all integrated with the Nexus terminal.
CVAT + Terminal Pipeline
CVAT projects with 5 PCB defect categories — solder_bridge, cold_solder, trace_defect, misaligned_component, missing_component. Terminal uploads PCB images and monitors GPU-accelerated inference in real time.
PCB Defect Detection Pipeline
Full CVAT project for EE-Design PCB Inspection — 5 defect categories with color-coded labels, issue tracking, and raw/constructor label editors. GPU-accelerated annotation and model training.
Automated Symbol Assembly
MAPO v3.0 gathers components, symbols, datasheets, and characterizations for your PCB designs. Analyze, gather, review — automated end-to-end.
228 Components. Fully Validated.
Symbol Assembly gathers 228 components, 228 symbols, 12 datasheets, and 221 characterizations automatically. Review results, re-analyze from scratch, or accept and continue to validation.
Configure Your Entire Stack
AI Connections (Claude, Cursor, Windsurf), provider routing via OpenRouter, local GPU compute for ML training, IDE plugin management, and bring-your-own-key support — all from one settings page.
Explore Every Capability
Real screenshots from production. Click any image to expand.
Compute Options: Multi-cloud GPU marketplace
HPC Dashboard: Cluster orchestration and job management
HPC Cluster Setup: SSH, Slurm, and container runtime configuration
Local Compute: Connect your own GPU with 3 commands
Start Building with GPU Compute
Download Nexus Forge — free, open source, MIT licensed. GPU marketplace, JupyterHub, HPC clusters, and terminal included.