sustainable-computing

SustainDC is a set of Python environments for Data Center simulation and control using Heterogeneous Multi Agent Reinforcement Learning. Includes customizable environments for workload scheduling, cooling optimization, and battery management, with integration into Gymnasium.

🔋🎯 Thread-level, NUMA-aware energy attribution model for multi-tenancy

An attempt to centralized some resources about sustainable programming, green computing and frugal engineering.

Estimates carbon emissions of compute jobs run on high-performance computing clusters.

Distributed LLM pretraining during renewable curtailment windows 🌱

Energy and carbon-aware extension of Vidur for sustainable LLM inference simulation and analysis 💡🌳

A logic programming solution to assess the ecological footprint of your ICT system

ECOLIFE: Carbon-Aware Serverless Function Scheduling for Sustainable Computing

A carbon monitoring dashboard for HPC to allow users to understand the environmental impacts of their jobs. Part of the Green Algorithms suite of tools.

Architectural blueprint and research specification for the Software-Defined Stochastic Inference Engine (SDSIE)—an adaptive runtime framework for low-energy AI inference on consumer hardware.

Carbon Aware cloud scheduler -- enabling a sustainable cloud

Reproducible, efficient, flexible batch processing using pipelines for sustainable software experiments

Sustainability Guidelines for Developers.

Green Anomaly Detection - Energy- and carbon-accounted anomaly detection: every model reports gCO2e alongside its detection score

A Green AI Knowledge Governance Engine for Enterprise RAG. Features FinOps-driven ingestion, PII redaction middleware, and strict lifecycle management to ensure zero-trust, cost-optimized AI. Built with FastAPI, React, Qdrant, and GPT-4o.

Self-powered, self-regulating neuromorphic unit: a LIF spiking network on a glucose biofuel cell, with a metabolic regulator that scales compute to available energy. Simulation + policy comparison (30 seeds, bootstrap CIs).

Reproduction package for FlexSure-Cert multi-risk certification of AI data-center demand response

Predict the energy consumption of Kubernetes workloads from their specs, before they run. Ground-truth energy labels from Kepler (eBPF) via Prometheus.

Package containing the core Green Algorithms methodology to estimate resource consumption and carbon footprint of computational workloads

Climate-aware forecasting and benchmarking for data centers