Optimization
Cost, latency, inference time, correctness, and energy trade-offs.
Publications
Research behind GAISSA.
Selected work behind the product direction: optimization, model serving, energy measurement, carbon evidence, and greener AI-based software systems.
Measurement
Energy labels, carbon evidence, utilization, and comparable reporting.
Green AI systems
Architecture and software-engineering decisions for efficient AI systems.
GAISSA-Optimizer · 2025 PROD 00236
Cost Optimization of Artificial Intelligence Systems with Sustainable Practices
Catalan technology transfer project · September 2025-March 2027
Selected work
Papers closest to the product story.
A focused selection explaining optimization, serving, energy labeling, architecture, and measurable sustainability evidence. The complete academic archive remains on the GAISSA research site.
Journal2025Computing
Impact of ML optimization tactics on greener pre-trained ML models
Evaluates how optimization tactics affect correctness, inference time, and energy consumption in pre-trained ML models.
Optimization tactics
Journal2025Journal of Systems and Software
Insights into resource utilization of code small language models serving
Studies serving code-oriented language models with different runtime engines and execution providers.
Model serving
Journal2025ACM TOSEM
Innovating for Tomorrow: The Convergence of Software Engineering and Green AI
Frames Green AI as a software engineering challenge, connecting research practice with operational impact.
Green AI agenda
Demo paper2024FSE Demo
GAISSALabel: A tool for energy labeling of ML models
A tool-centered publication around measuring and communicating ML model energy efficiency in a practical way.
Energy labeling
Conference2024CAIN at ICSE
Identifying architectural design decisions for achieving green ML serving
Connects greener ML serving with architectural decisions, deployment choices, and measurable efficiency outcomes.
Architecture
Conference2024MSR
Analyzing the Evolution and Maintenance of ML Models on Hugging Face
Repository mining study about how ML models evolve, change, and are maintained in the Hugging Face ecosystem.
Model evolution
Conference2023ESEM
Exploring the Carbon Footprint of Hugging Face's ML Models
Empirical work measuring and analyzing carbon-footprint information across model repositories.
Carbon evidence
Conference2023SEAA
Towards green AI-based software systems: an architecture-centric approach (GAISSA)
The research foundation behind GAISSA: architecture-centric methods for greener AI-based software systems.
GAISSA foundation
Research context
Continue into the complete research line.
GAISSA UPC research project
Architecture-centric research for greener AI-based software systems.
Full GAISSA publication archive
The complete UPC-maintained list of papers, theses, seminars, and related work.
GAISSA-Optimizer project
Technology transfer connecting the research line with product and market value.