The Problem

Most personal computers sit idle 90% of the time. Meanwhile, ML training and gaming workloads cost a fortune on cloud GPU instances. I wanted to bridge that gap — turn idle hardware into useful compute.

What I Built

ComputePool — a hub-and-spoke distributed compute grid. Your Zo Computer acts as the control plane. Idle laptops and PCs become worker nodes that poll for jobs, execute workloads, and earn credits.

Architecture

Node Agent (Python) ← polls → Hub API ← dispatches → Worker Pool ↓ Credit Ledger ↓ Cashout System

Enter fullscreen mode

Exit fullscreen mode

Node Agent ( node-agent/node_agent.py ):

Polls hub every 30s for available jobs

Reports GPU tier (RTX 4090 = 3x credit multiplier)

Streams results back on completion

Hub ( hub/hub.ts ):

FastAPI backend on Railway

Job queue with priority based on GPU tier

Credit ledger per node

Regional multipliers (Indian region: 0.7x)

Dashboard ( frontend/ ):

Next.js 14 on Vercel

Real-time job status, credit balance, node…