HANSAL: Hybrid AI for Next-gen: Sustainable, Affordable, and Lightweight

Introduction

Recent advances in AI demand significant computational resources, which presents cost and sustainability barriers for small and medium-sized enterprises (SMEs).

Project HANSAL addresses these challenges, offering a benchmarking framework that balances accuracy, runtime, cost, and energy consumption for a DL solution.

This project is a benchmarking framework that is also featured in the Master's Thesis @ HAW Kiel:

Study on Proof-of-Concept Benchmarking Framework for Resource-Optimised AI: Efficient Performance, Cost, and Sustainability for Small-Medium Enterprises (SMEs)

Project Scope

• Exploration and testing of AI benchmarking tools suitable for SMEs.
• Evaluation of currently available hardware architectures (GPUs, AI accelerators) for affordability and scalability.
• Implementation of a lightweight benchmarking workflow using open-source libraries (e.g. CodeCarbon, Zeus, Perun).
• Creation of an open-source repository for replicable benchmarking experiments, supporting business case-specific needs.

Project Objectives

• Provide SMEs with the tools and know-how to adopt AI cost-efficiently and sustainably.
• Serve as both an academic contribution and an actionable industry guide.
• Structure data pipelines for transparent performance, allowing resource optimization tailored to SMEs.