FREN

Noctua Polaris, independent AI & data consulting

Independent AI & Data Consulting

Assessment, bespoke models, validation readiness and ongoing support for SMEs, industry and aviation.

Get in touch

Services

You can start with any of them, or combine them.

AI & Data Assessment

A clear view of what is worth doing, with or without AI, and where to start.

Who it’s for

SMEs and mid-sized companies wondering where to start with AI. Industrial firms considering a model that want to check it is worth it before investing.

Read the description

A few days of workshops with your teams to understand your business, tools and data. You leave with a prioritised list of actions, based on their expected benefits and implementation costs. AI is only recommended when its expected benefits justify the investment.

Bespoke AI Models

A model tailored to your problem, with measured performance and identified limitations.

Who it’s for

SMEs, mid-sized companies and start-ups with a forecasting or optimisation problem and no in-house data scientist. Start-ups and SMEs running an innovation project. Organisations developing a data service over several years.

Read the description

For operational needs such as demand forecasting, maintenance planning or resource allocation, the model is built using your data and integrated into your existing tools. For innovation projects, the work draws on relevant research to develop an initial tested version, designed with production deployment in mind. In both cases, the model’s performance is assessed overall and across different situations to identify the conditions in which its results are less reliable.

AI Validation Readiness

Designing models to meet validation requirements from the outset, or building the technical evidence needed to demonstrate the reliability of existing models.

Who it’s for

Aerospace, space and defence suppliers that must demonstrate their models’ reliability to their prime contractors. Teams putting a predictive model on structured data into production under reliability requirements.

Read the description

If you are starting a project, the work begins with the requirements set by your prime contractor or applicable regulations, and the model and its data pipeline are designed to address them from the outset. If you already have a model, the assessment covers the entire chain, from data collection through to the model in service, to identify gaps against those requirements. The solutions are then built with your team, with each step tested against your data. In aviation and other safety-critical systems, this helps prepare the model’s assurance case. I do not issue certification.

Fractional AI Lead

AI projects well-scoped and kept on track, without hiring a full-time AI lead.

Who it’s for

SMEs and mid-sized companies running AI projects without a dedicated AI lead, often relying on someone working on AI part-time and without specialist support. Tech start-ups looking for senior guidance without hiring.

Read the description

A few days a month, you get senior guidance to scope your AI projects, assess technical decisions and support your team or external providers. When selecting tools or suppliers, you benefit from independent advice and help drafting specifications. The person driving AI initiatives in your company has experienced support to rely on. The decisions remain yours.

Get in touch

Experience

A few projects from my career.

Air traffic management

The challenge was twofold: landing more aircraft per hour without reducing safety, and demonstrating to the regulator that the AI models used were reliable. For around ten years, I have worked on these machine learning models, which compute the safety buffers between aircraft, and on how to demonstrate their reliability: an error target met for each type of aircraft pair and each weather condition, and fallback to the separation rule in force when the model is not reliable enough. The system has been tested at three major European airports, in real-time simulations and shadow-mode trials with air traffic controllers. It is one of the use cases in EASA’s Concept Paper, was presented at the EASA AI Days 2026 as part of the AI Olympics, and the work has been published (DASC 2018, SESAR Innovation Days 2022).

Offshore shipping

The shipowner wanted to improve how crews were organised and how fleet maintenance was managed. Using its operational data, I measured the actual workload on board, then designed a tool to prioritise maintenance tasks, based on heuristics and operations research. The workload analysis informed crew organisation, and the tool was rolled out across around ten vessels, refined over several iterations and eventually integrated by the client into its own system.

Air passenger demand forecasting

Airports need to plan their staff and resources weeks ahead without knowing how many passengers each flight will carry. I design and develop a forecasting service that addresses this need, from the data pipelines to the machine learning models, which are put into service only if they outperform a simple baseline forecast. The service runs daily in production, with automatic monitoring of its accuracy, and forecasts the expected number of passengers on each flight, up to eight weeks ahead.

Assessments

Around ten AI and data assessments for SMEs and mid-sized companies (IT services, medical event management, spatial planning, landscaping, logistics, mobility), each ending with a prioritised action roadmap, plus a data and AI master plan for a major infrastructure operator.

About

Guillaume Stempfel

Guillaume Stempfel

  • PhD in AI
  • 15 years in AI and data consulting
  • 10 years of aviation projects

After completing my PhD in machine learning, I took on technical leadership roles in AI projects across several sectors, including air traffic management. I also led a team of around twenty people and contributed to European standardisation work on AI in aviation. Since 2026, I have worked independently, taking on a small number of engagements at a time and delivering them myself.

Selected publications

All publications

Contact

contact@noctua-polaris.eu

+33 4 84 89 71 81 LinkedIn

96 rue Paradis, 13006 Marseille, France