The client received a production-ready testbed to explore LLM performance across a range of ML tasks.
Meet our client
Client:
Industry:
Market:
Technology:
Client’s Challenge
A major EU institution needed a fast and flexible way to test open-source LLMs across various machine learning tasks. With a constrained budget and timeline, their internal teams were unable to build such a platform independently.
Our Solution
We developed a lightweight, modular framework in Python, using Open AI API standard, that allows users to describe ML problems in natural language, select prompt strategies and model combinations, and automatically test all viable setups. The tool enables exporting the best-performing configuration and works exclusively with open-source LLMs. We also deployed several models with TGI to support internal benchmarking.
Client’s Benefits
The client received a production-ready testbed to explore LLM performance across a range of ML tasks. Internal teams could now benchmark, compare, and reuse configurations. The solution also supported the EU’s focus on open and transparent AI.
