Research and development

Research is how we stay ahead, and how our clients benefit from the most recent advances in data and AI before they become mainstream.

Why research matters at Emerton Data

Data and AI evolve faster than any other field of consulting. Methods that were experimental two years ago are now in production, and the questions our clients ask, on generative and agentic AI, optimisation or the reliability of models, rarely have an off-the-shelf answer. Advising them well requires understanding these techniques from the inside.

We test new methods ourselves and turn the results into tools, products and papers.

Our team includes several PhDs, we work closely with leading academic laboratories, and we publish our work. This combination of scientific depth and business experience is what allows us to make complex subjects simple without making them simplistic.

Our research programme

Solutions built on the advanced use of data, artificial intelligence, advanced algorithms and scientific optimisation methods require substantial research and development.

Emerton Data has launched research on key topics for today's and tomorrow's economy. This work supports both our consulting activities and the development of innovative solutions. We have initiated major projects such as the development of methods for responsible AI, including the interpretability and fairness of algorithms.

Our research takes shape through

  • Innovative solutionsLaunch of products and tools embedding our own algorithms.
  • Scientific papersPapers submitted in the field of machine learning.
  • Scientific Excellence PrizeAwarded in the AI for Industry challenge organised by the Île-de-France Region, by a jury of renowned academic researchers.
  • Academic partnershipsJoint work with leading research laboratories.

Academic partners

Open-source research / 2026

GLIDE

GLIDE is an open-source Python library that brings statistical rigor to the evaluation of GenAI and agentic systems.

View the code on GitHub Read the paper

We help teams move past the false choice between cheap, biased LLM-judge scores and accurate but costly human review.

GLIDE combines a small set of human-labeled examples with large-scale proxy annotations, using prediction-powered inference to correct for the judge's bias rather than ignore it.

The result is unbiased performance estimates with formal confidence intervals, spanning the full pipeline from optimal sampling and annotation to estimation and continuous production monitoring.

Built on peer-reviewed statistical methods and released as open source, GLIDE gives teams metrics they can defend to their peers, their stakeholders, and themselves.

Paper Industrializing Prediction-Powered Inference: The GLIDE Library for Reliable GenAI and Agentic Systems Evaluation
Grégoire Martinon, Ibrahim Merad, Mohamed Raki

From research to products

Alongside our consulting work, we develop innovative deeptech products through our start-up studio, Kaukana Ventures.

PhDs in our team

The Emerton Data team includes several PhDs. Examples of profiles:

Aimé Lachapelle Ph.D. in Applied Mathematics Romain Aeberhardt Ph.D. in Economics Grégoire Martinon Ph.D. in Physics Thibault Cuisiniere Ph.D. in Microbiology and Immunology Pierre Six Ph.D. in Mathematics and Control Theory Ibrahim Merad Ph.D. in Applied Mathematics David Thery Ph.D. in Computer Science

Research insights

All news

Join us

Put your data and AI expertise to work on major international business challenges. Join our team and help shape the future of industries worldwide.

See careers

Work with our research team