CydroFlux AI

CydroFlux AI implements an end-to-end, sustainability-aware routing engine for AI workloads. It collects and normalises telemetry for six data-centre regions, scores each facility with a transparent Green Intelligence Score, enforces latency, capacity and residency guardrails, chooses a facility by a deterministic hierarchy and records the full reasoning. A language model helps interpret natural-language requests but is confined to intent. . At baseline the facility ranking is robust to the weights and to telemetry noise, and the latency guardrail is what determines the winner for a given client. The contribution is the orchestration and its inspectability, not a new optimisation algorithm.

[1]     E. Masanet, A. Shehabi, N. Lei, S. Smith and J. Koomey, “Recalibrating global data center energy-use estimates,” Science, vol. 367, no. 6481, 2020. [2]     D. Patterson et al., “Carbon emissions and large neural network training,” arXiv:2104.10350, 2021. [3]     P. Li, J. Yang, M. A. Islam and S. Ren, “Making AI less ‘thirsty’: Uncovering and addressing the secret water footprint of AI models,” arXiv:2304.03271, 2023. [4]     A. Radovanović et al., “Carbon-aware computing for datacenters,” IEEE Trans. Power Syst., 2023. [5]     The Green Grid, “Water usage effectiveness (WUE): A Green Grid data center sustainability metric,” White Paper, 2011. [6]     L. A. Barroso and U. Hölzle, “The case for energy-proportional computing,” IEEE Computer, vol. 40, no. 12, 2007. [7]     Gemma Team, Google DeepMind, “Gemma: Open models based on Gemini research and technology,” arXiv:2403.08295, 2024. [8]     C. Percival, “Stronger key derivation via sequential memory-hard functions,” BSDCan, 2009. [9]     OWASP Foundation, Password Storage Cheat Sheet. [Online]. Available: https://cheatsheetseries.owasp.org/ [10]  M. Jones, J. Bradley and N. Sakimura, “JSON Web Token (JWT),” RFC 7519, IETF, 2015.