Built for
GPU-intensive infrastructure.
HexaGrid integrates live wholesale electricity markets, neural price forecasting, reinforcement learning dispatch, and quantum-ready optimization into a unified control layer — designed for AI data centers and multi-site GPU fleets.
Three-tier modular architecture
Each tier is independently deployable and communicates exclusively over a versioned REST API. This lets you scale any individual component — the RL agent, the fleet orchestrator, the GPU monitor — without touching the rest of the platform.
Ten Python modules each own their own data persistence via SQLite (PostgreSQL + TimescaleDB for production scale). A single FastAPI application exposes all modules through one versioned interface with auto-generated OpenAPI docs. The dashboard is a static HTML file — no build step, no dependencies.
How HexaGrid is built
Modular, not monolithic
Every pillar is an independent Python module with its own data persistence and API surface. Components deploy, update, and scale in isolation without disrupting the rest of the platform.
Live data, not stale models
HexaGrid maintains live connections to EIA and Electricity Maps continuously. Every dispatch decision runs on current market state — not monthly averages, estimates, or historical proxies.
Quantum-ready by design
The QAOA scheduler runs on Cirq's hardware abstraction layer. The same circuit architecture migrates to real QPU backends as quantum hardware matures — with zero code changes required.
Full operator visibility
Every routing decision, RL action, and hardware alert is logged and visible in the dashboard. Operators see the reasoning behind every recommendation and retain full override capability at all times.
Zero build dependencies
The dashboard is a single static HTML file. No webpack, no Node, no build pipeline. It serves from nginx, an S3 bucket, or a local Python server — operational in under 60 seconds from a fresh clone.
Continuous self-improvement
The PPO agent and IsolationForest anomaly detector retrain on live data. HexaGrid doesn't degrade over time — it improves as market patterns and hardware behavior accumulate in its training data.
What HexaGrid runs on
| Component | Technology | Purpose | Version |
|---|---|---|---|
| Digital Twin | SimPy discrete-event simulation | GPU rack power flow, PUE modeling, thermal dynamics | 4.1+ |
| LSTM Forecaster | TensorFlow LSTM | 120-minute grid price prediction, real-time inference | TF 2.13+ |
| QAOA Scheduler | Cirq + TensorFlow Quantum | Quantum-classical hybrid job scheduling | Cirq 1.2+ |
| RL Dispatch Agent | Stable Baselines 3 PPO | Adaptive workload deferral, 200k training steps | SB3 2.0+ |
| API Platform | FastAPI + Uvicorn | Versioned REST, OpenAPI /docs, CORS, structured logging | 0.104+ / 0.24+ |
| Carbon Layer | Electricity Maps API | Live gCO₂eq/kWh per ISO, Pareto cost–carbon analysis | v3 |
| Fleet Orchestrator | Python + SQLite | Multi-site composite scoring, routing history, 4 strategies | Python 3.11+ |
| GPU Monitor | pynvml + scikit-learn | NVML telemetry, IsolationForest anomaly detection, health scoring | 11.5+ / 1.3+ |
| GPU Acceleration | CUDA + NVIDIA Driver | TF training, RL optimization, quantum circuit simulation | CUDA 12.x / 525+ |
Quantum Clarity LLC
HexaGrid is built by Quantum Clarity LLC, a deep-technology firm working at the intersection of quantum computing, reinforcement learning, and real-time energy systems.
The platform is fully operational at v1.0.0. Every benchmark was produced on real NVIDIA GPU hardware with live EIA and Electricity Maps data feeds — not simulated environments or theoretical projections.
Quantum Clarity also develops the QLLMe™ hybrid optimization engine and the QuantaCore™ modular quantum computing platform, which provides HexaGrid's long-term QPU migration path as quantum hardware matures.
Production-validated at v1.0.0. Deployed on bare metal, WSL2, and AWS EC2. All results from live infrastructure — not simulations.
Contact us
Interested in working together? Fill out some info and we will be in touch shortly. We can’t wait to hear from you!