HarmonE: A Self-Adaptive Approach to Architecting Sustainable MLOps
Hiya Bhatt, Shaunak Biswas *, Srinivasan Rakhunathan, Karthik Vaidhyanathan · 2025
Read HarmonE: A Self-Adaptive Approach to Architecting Sustainable MLOps by Hiya Bhatt, Shaunak Biswas *, Srinivasan Rakhunathan, Karthik Vaidhyanathan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-13693.txt.
Description
Machine learning systems rot quietly: data drifts, models stale, and retraining burns energy and money. Hiya Bhatt and co-authors present HarmonE in this 2025 paper, a design that lets a learning pipeline watch and repair itself with sustainability in mind. Architects set goals and thresholds up front, covering accuracy, energy use, and data shift.
A monitor-analyze-plan-execute loop then tracks those measures at run time and fires fixes when limits are crossed. The team tested the design on a digital twin of a traffic prediction system. The twin let them replay traffic shifts safely and watch the pipeline adapt. HarmonE held accuracy while respecting the set goals, showing a route to greener, longer-lived learning systems.
Catalog record
| CREATOR | Hiya Bhatt, Shaunak Biswas *, Srinivasan Rakhunathan, Karthik Vaidhyanathan |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-13693