Home / R&D Lab

We help drive the AI revolution through our R&D lab

Novelis Research, our in-house R&D lab of PhDs and AI researchers, is part of our DNA. We invest 25% of our revenue in fundamental research on the core principles of AI and generative AI.

Our commitment

One foot in the future

At Novelis, we are committed to using new technologies as tools to better address our clients' operational challenges and support their transformation. To meet these needs, we built an ambitious R&D lab, backed by significant investment.

LeBrain, the enterprise AI platform we publish, was born in this lab.

“We want to keep one foot in the future and help build it by investing in scientific research and in the technologies of tomorrow. Innovation is part of Novelis's DNA.”

El Hassane Ettifouri, Partner and Chief Innovation Officer

Our research areas

Three visions that guide the lab

Artificial intelligence
AI plays a leading role in how organizations operate digitally. Understanding and mastering it, delivering AI projects, and using it to meet business needs all take time to learn. We have been doing this work since we were founded.
Intelligent automation
RPA (robotic process automation) covers the tools and techniques used to automate business processes. Novelis is an RPA pioneer and uses intelligent automation in its own operations.
Software engineering
AI is used in many fields, from medicine to automotive to manufacturing. Recent years have brought major advances in AI and software engineering. We want to push science forward toward artificial developers.

Scientific publications

The lab publishes in the scientific press

The Novelis R&D lab helps advance AI research worldwide. All articles are written in English.

July 2024 · Multimodal AI
Graphical User Interface Agents Optimization for Visual Instruction Grounding Using Multi-Modal Artificial Intelligence Systems
Published on arXiv and submitted to the journal Engineering Applications of Artificial Intelligence. Introduces SIC (Search Instruction Coordinates), a multimodal solution that takes a natural language instruction and a screenshot, and locates the interface component where the instruction should run.
June 2024 · Language models
Benchmarking Open-Source Language Models for Efficient Question Answering in Industrial Applications
Published on arXiv and submitted to the journal Engineering Applications of Artificial Intelligence. Compares open-source and proprietary models on question answering, to identify lightweight models that run on a standard CPU with comparable performance.
April 2024 · Code generation
Low-Cost Language Models: Survey and Performance Evaluation on Python Code Generation
Published on arXiv and submitted to the journal Engineering Applications of Artificial Intelligence. Evaluates models that run on a standard CPU, proposes a step-by-step reasoning prompt and a set of 60 programming problems.
September 2023 · LLM reasoning
GPT-3.5, GPT-4, or BARD? Evaluating LLMs Reasoning Ability in Zero-Shot Learning and Performance Boosting Through Prompts
Published by Elsevier (ScienceDirect). A technical evaluation across eleven reasoning datasets, with prompts that improve the performance of the models studied.
May 2023 · LLM reasoning
GPT-3.5 vs GPT-4: Evaluating ChatGPT’s Reasoning Performance in Zero-shot Learning
Published on arXiv. First version of the study above, across eleven reasoning datasets.
May 2023 · Code generation
A Comprehensive Review of State-of-The-Art Methods for Java Code Generation from Natural Language Text
Published by Elsevier (ScienceDirect). A review of how deep learning models for Java code generation have evolved, along with their datasets and metrics.
February 2023 · Code generation
JaCoText: A Pretrained Model for Java Code-Text Generation
World Academy of Science, Engineering and Technology, presented at the International Conference on Code Generation and Implementation (Paris, October 2022).
June 2021 · SQL
SQL Generation from Natural Language: A Sequence-to-Sequence Model Powered by the Transformers Architecture and Association Rules
Journal of Computer Science, volume 17, no. 5, 2021. Sets a new state of the art on the WikiSQL dataset for weakly supervised SQL generation.
August 2020 · SQL
Artificial Neural Networks for Text-to-SQL Task: State of the Art
International Conference on Smart Information and Communication Technologies, Springer.
August 2020 · Code generation
NL2Code: A Corpus and Semantic Parser for Natural Language to Code
International Conference on Smart Information and Communication Technologies, Springer. Long-term goal: build an application from a specification written in natural language.
July 2020 · SQL
SQL Generation from Natural Language Using Supervised Learning and Recurrent Neural Networks
International Conference on Artificial Intelligence and Industrial Applications, Springer.
December 2019 · SQL
Text2SQLNet: Syntax Type-Aware Tree Networks for Text-to-SQL
Europe, Middle East and North Africa International Conference on Information Systems and Technologies to Support Learning, Springer.

White paper · May 12, 2026

From RPA to governed agentic AI

Automation has entered a new phase. After RPA and traditional workflows, companies now have to handle processes that are more variable, more document-heavy, and more dependent on business context. The real challenge now is scale: governance, security, auditability, integration with IT systems, and performance management.

In this white paper, our AI R&D Lab shares a practical approach to moving from AI prototypes to reliable, supervised systems that can run at scale.

Get the white paper

International rankings

Recognized alongside AI leaders

The lab's work on generating code and queries from natural language has been recognized in international research challenges.

1stCodeXGLUE challenge, run by Microsoft, December 2021
2ndSPIDER challenge, run by Yale University, March 2021
7thWikiSQL challenge, run by Cornell University, March 2021

Have a project or a question?

Our researchers and experts are ready to answer your questions.