The Problem: The GPU Paradox

Currently, the compute market is broken for two major groups:

• Institutions: Large-scale hardware owners (data centers, universities) leave roughly $2.2M per year on the table (per 100 idle GPUs) because their hardware sits unused. They cannot join decentralized networks due to a lack of compliance, KYC, and geo-fencing controls.
• Startups & AI Devs: Teams are wasting $1.2M per year (per 100 GPUs) overpaying for centralized cloud providers. While cheaper decentralized alternatives (DePIN) exist, viable options are fragmented across numerous networks, each with its own complex CLI, token requirements, and configuration specs.

The Solution: Necto

Necto is a Two-Sided Institutional Compute Marketplace that bridges the gap between idle enterprise hardware and the decentralized compute economy.

What people use it for:

• Frictionless AI Deployment: Instead of manual configuration, users simply describe their task (e.g., "I need to train a PyTorch model") to an AI Agent. Necto automatically infers the required Docker images, VRAM, and CPU specs.
• Cost Optimization: Necto acts as the "Expedia for GPUs," searching across 8 major networks (Akash, Render, io.net, etc.) to find the cheapest, most compliant provider in real-time.
• Monetizing Idle Hardware: Institutions can safely list their "dark" compute to earn revenue, using Necto’s operator dashboard to toggle machines and monitor earnings instantly.

How it makes tasks easier & safer:

• Enterprise-Grade Controls: Unlike raw DePIN protocols, Necto allows institutions to enforce KYC requirements, Geo-fencing (e.g., "only deploy in the EU"), and Workload Blocking to ensure regulatory compliance.
• Unified On-Chain Settlement: Users don't need to hold 8 different tokens. Payments are handled via USDC escrow on ADI Chain, which manages the settlement and provides a verifiable audit trail for accounting.
• Visual Automation: The Workflow Builder allows advanced users to create "Compute Pipelines." You can set a budget guard (to prevent overspending) and a compliance gate, then automatically deploy to multiple networks via a simple drag-and-drop interface.
• Complete Traceability: Every deployment, configuration change, and payment is logged in a transparent Audit Log, making decentralized compute finally "audit-ready" for professional teams.

Challenges: One major struggle point was building the AI agent at the core of the system. Picking the right libraries, building a robust system of tools to scan providers, read user input, parse to correct SDL and interface with provider's directly was a lot of tools and orchestration to get ready for demos.

We explored different options, like Anthropic's Ai sdk, google's agent development kit, before eventually settling into Vercel's ai-sdk. Every integration helped it utilize tools even better than the last, until it's at a point where we're very happy with it. Took quite a bit of plumbing to finish.