Open source consulting Paris · Remote · On-site

Data science consulting
from analysis to deployment.

We help organisations analyse their data, develop models and deploy applications. We mainly use open-source tools and work within your existing technical environment.

Discuss your project → Our training courses
our_process.sh copy 
# typical project stages

$ project_workflow
✔ scope defined  # objectives and constraints
✔ data reviewed  # quality and availability
✔ model developed # Python, R or LLMs as needed
✔ deployed       # API, application or internal service
✔ documented     # documentation and knowledge transfer

$ echo "deliverables: code, documentation, handover"

our process

How we work

Our approach is adapted to the project, from the initial assessment to delivery and documentation.

01
🎯

Project scoping

We discuss your objectives, the decisions the analysis should support, the available data and any technical or organisational constraints. This helps us agree on a realistic scope.

What to build? What data exists? Expected benefits? What's the risk?
02
🔍

Analysis and modelling

We explore the data, prepare variables, compare methods and validate the results. Depending on the project, we use Python (scikit-learn, PyTorch, statsmodels) or R (tidymodels, brms). We can use MLflow to track experiments.

Key findings Model selection Validation
03
⚙️

Development and deployment

Where deployment is needed, we develop maintainable applications and automate testing and delivery. Depending on your infrastructure, we may use uv, GitLab CI/CD, Docker, FastAPI, Streamlit or Dash. Deployment can take place on your servers or in your cloud environment.

FastAPI / Docker GitLab CI/CD On-premise
04
🎓

Documentation and handover

We provide documentation and review the code with your team. Training and support can be included to help your staff maintain and develop the solution.

Documentation Team training Handover

services

Our services

We develop data analysis tools, models and applications using technologies suited to your needs.

🤖
AI & LLM systems
We develop RAG applications for internal documents, adapt open language models and build agents that use external tools. On-premise deployment is possible when data needs to remain within your infrastructure. We take data governance, traceability and applicable EU AI Act requirements into account.
RAGMistral AILlamaLangChainLlamaIndexon-premiseAI Act
→
⚙️
Python ML / MLOps
We develop machine learning models and help integrate them into existing applications. Depending on the project, we use uv for dependencies, MLflow for experiment tracking, FastAPI for APIs, Docker for deployment and GitLab CI for automated tests.
uvGitLab CI/CDFastAPIMLflowscikit-learnPyTorchDocker
→
📈
Data apps & dashboards
We build interactive applications with Streamlit, Dash and Shiny. These can range from internal prototypes to applications with authentication, deployed on your infrastructure.
StreamlitDashShinyPlotlyggplot2QuartoDocker
→
📊
R & statistical modelling
Statistical modelling, econometrics, survival analysis, time-series forecasting and text analysis. We also support SAS-to-R migrations, including checks on results, and help configure Posit tools such as Workbench and Connect.
RtidymodelsbrmsforecastrenvPositSAS migration
→
🔬
On-demand data analysis
We carry out targeted analyses using your data, including segmentation, text analysis, forecasting, scoring and experimental design. Typical turnaround is 3 to 5 working days, depending on the scope and data provided. A time-credit option is available.
NLPforecastingsegmentationscoringDOEsurvival
→

open source stack

Our main tools

We favour open-source tools that make projects easier to maintain, document and transfer to your teams.

🐍Python
ML & AI
scikit-learnPyTorchHuggingFaceLangChainLlamaIndexOllamavLLM
📦Python
Packaging & quality
uvruffmypypytestpre-commit
📊R
Stats & viz
tidyversetidymodelsggplot2brmsforecastrenv
🔧DevOps
CI/CD & deployment
GitLab CI/CDDockerMLflowDVCFastAPIPrefect
🖥️Apps
Dashboards & apps
StreamlitDashShinyPlotlyQuarto
🏗️Infra
R environment
RStudio ServerPosit WorkbenchPosit ConnectPositron

philosophy

Why we favour open source

Open-source tools give your teams more flexibility in how they maintain and develop their projects.

Flexibility
Code and documentation
We aim to deliver code and documentation your team can reuse and extend. Open-source components can also make it easier to change tools or infrastructure, subject to their licences and dependencies.
Auditability
Traceability and review
Open-source tools can help teams review implementations and document how a system works. Regulatory compliance still depends on the specific system, its use and the controls put in place.
Data sovereignty
Deployment options
We can deploy applications on your own servers or in a private cloud. For projects involving sensitive data, we discuss hosting, access controls and any external services before deployment.
Reproducibility
Reproducible environments
We use tools such as uv, renv and Docker to make development environments easier to reproduce and maintain. Automated tests help identify differences between environments.

track record

Experience across different sectors

We have worked with organisations in energy, finance, healthcare, retail, research and telecommunications.

EDFEngie AXADior L'OréalOrange LCLOCDE UbisoftNestlé ThalesScania CNRSCerba Healthcare CofaceInstitut Curie Société GénéraleIfremer

Tell us about your project.

Describe your project, the data available and any constraints. We aim to reply within 48 hours to discuss the scope and possible next steps.

Phone +33 1 72 25 40 82 Email info@stat4decision.com Response within 48h
Get in touch → ← Back to home