All projects
Software & Agents
SustainML Framework
An EU framework that guides ML development toward lower-energy, lower-CO₂ models.
eProsima · Horizon Europe SustainML project
Overview
SustainML helps developers describe a problem and get model, hardware and carbon-footprint recommendations across the ML lifecycle. As lead backend developer at eProsima, I built the C++ core that connects the framework's modules over Fast DDS. It is exposed to Python through SWIG, with a Qt/QML frontend and REST APIs.
With European partners I integrated transformer-based modules and RAG pipelines, including a knowledge graph built from the Hugging Face model hub and queried with SPARQL to ground model suggestions.
Highlights
- 01Distributed module architecture over DDS with a Python-friendly API.
- 02Qt/QML desktop frontend and REST service for remote use.
- 03Hugging Face → knowledge graph crawler with LLM-backed conversational search.