Nexa Core
Decentralized GPU Compute Marketplace, Multi-Role Vendor Onboarding & Real-Time WebSocket Telemetry
Dedicated Enterprise GPU Capacity Portal
Contract-first reserved GPU infrastructure platform engineered for enterprise ML teams.
Real-Time WebSocket Stream, Vendor Portal & Allocation Plans
Nexa Core GPU Marketplace & WebSocket Telemetry Suite
Next.js 16Decentralized GPU marketplace for hardware vendor onboarding, client compute allocations, dynamic multi-role RBAC, and sub-second real-time WebSocket telemetry streaming.
Cluster 01 (US-East)
● Active AllocationWorkload: Llama-3.3-70B Fine-Tuning · Location: Ashburn, VA
01.System Overview & The Engineering Challenge
Nexa Core is a decentralized GPU cloud computing and hardware marketplace started in 2025 and currently in active development. The platform allows hardware providers to register as vendors and monetize idle GPU clusters (H100, A100, RTX 4090), while AI startups and machine learning researchers can request customized GPU allocation plans, complete payments, and gain immediate secure access. The platform features dynamic multi-role access control, a Super Admin telemetry control plane, and real-time WebSocket streams delivering sub-second GPU metrics including VRAM utilization, temperature, power draw, and compute load.
Building a high-throughput GPU rental marketplace requires coordinating asynchronous hardware provisioning, multi-role security boundaries with dynamic access permissions, transaction-safe allocation agreements, and high-frequency WebSocket telemetry streams handling live hardware heartbeats without degrading database performance.
02.Ali's Specific Role & Deliverables
Ali is architecting the full-stack system using Next.js 16 for the client application and operator portal, Node.js for backend orchestration, PostgreSQL for multi-tenant state and allocation contracts, and WebSockets for real-time live GPU telemetry. He designed the vendor hardware listing portal, client allocation request workflow, dynamic multi-role permission system, and the live WebSocket hardware monitoring dashboard.
Key Technical Contributions:
03.Architectural Highlights & Decisions
Dynamic Multi-Role RBAC & Access Matrix
Constructed flexible authorization layers supporting dynamic role generation, per-resource permission policies, and secure multi-tenant isolation across Super Admins, hardware vendors, and compute clients.
Real-Time WebSocket GPU Telemetry Engine
Engineered high-frequency WebSocket streaming pipelines broadcasting sub-second hardware metrics (VRAM utilization, thermal readings, power draw in Watts, and active container instances) from distributed vendor nodes.
Vendor Hardware Ingestion & Client Allocation Engine
Designed automated capacity scheduling connecting client compute requests with verified vendor GPU nodes, featuring automated billing reconciliation and instant SSH/API key issuance.
Super Admin Global Fleet Monitoring Control Plane
Built full-scale operations dashboards providing unified visibility across total active cluster FLOPS, vendor payout logs, allocation contract health, and infrastructure availability.