Pertimm - LLM, Machine Learning and Semantic Search Expertise

We are a French technology company built around natural language processing and information retrieval. Our core work covers large language models and machine learning applied to real operational problems: retrieval augmented generation, semantic and vector search, embedding models, fine tuning and adaptation of open models to specific domains and languages, named entity recognition, document understanding, and matching engines that align unstructured text against structured criteria at scale.

What distinguishes our contribution to a consortium is the engineering behind the models. We build search and inference systems that run on realistic infrastructure, respond within production latency budgets, handle multilingual corpora, and remain explainable enough for regulated environments. We work on data quality, evaluation methodology and model monitoring as seriously as on model selection, which matters when a project has to demonstrate results rather than describe them.

We are active in the European research landscape, currently contributing to the OPEVA project, and we collaborate continuously across both EU and French funding programmes, including Horizon Europe and the France 2030 rounds. Working on both sides gives us a practical understanding of how each programme evaluates technical ambition, and how to position a work package so it is credible to reviewers.

Over the years we have built a wide network spanning industrial groups, innovative SMEs, large enterprises and academic partners. That network lets us suggest complementary participants and help assemble consortia that are balanced in scale, geography and capability.

We are looking to meet organisations preparing proposals where language models, applied machine learning or intelligent search form part of the technical core, or where they can strengthen the exploitation and market uptake strategy. If you are building a consortium or filling a specific technical gap, I would be glad to discuss how Pertimm could contribute.