📊 Senior Data Scientist: Industrial AI Platform

🏢 The Company

An early-stage startup building a machine learning platform for industry, backed by a software vendor that is already well-established in the field.

• The product: predictive maintenance, quality correlation, energy monitoring, and OEE optimization. Two families of algorithms, four use cases on the production line.
• The promise: that the factory will receive actionable recommendations without having to hire a data scientist. The entire ML workflow (variable selection, preprocessing, model selection, calibration, and drift detection) is automated.
• Clients: world-class manufacturers in the automotive, aerospace, and food industries.
• The data team: It doesn't exist yet. You're the first one.

🎯 The Role

You've been hired as a senior data scientist, and the role is intended to evolve into a Head of Data position. In practice: first, you produce results; then, you establish a structure; and finally, you hire staff.

• You design and scale up the models that are at the heart of the product: anomaly detection, correlation between process parameters and scrap, and energy consumption modeling.
• You put these models into production and ensure they stand the test of time: drift, re-training, reliability.
• You build the data pipelines and the associated architecture as new industry sources become available.
• You establish the governance framework: quality, consistency, documentation, monitoring, and access.
• You translate the results into shop floor language. A process engineer must be able to act on your model output without an interpreter.
• Eventually, you'll build and lead the data team.

⚠️ What You Need to Know

The scope of the role is broader than the job title suggests. Right now, it’s a senior data scientist position on a team that doesn’t yet exist: you’ll be handling data engineering, MLOps, and modeling all on your own. The Head of Data role is being developed as the company grows; it’s not guaranteed upon signing the contract.

Another key feature: the product’s value doesn’t lie in the model itself, but in what surrounds it. The customer doesn’t have a data scientist—and never will. The real challenge is to automate everything a data scientist would do manually, and to make it reliable when applied to factory data that varies from one site to another.

🔨 Technical Challenges

• Without labeled data: factories do not label their defects. The models must learn what normal behavior looks like and flag any deviations.
• A complex issue: hundreds of process variables per line, only a handful of which actually matter. Identify the right ones and explain why.
• Mandatory explainability: An unjustified recommendation is not implemented in the workshop. Interpretability is not a bonus; it is a prerequisite for adoption.
• Industrialization: The models are running for hundreds of customers, processing diverse data sets, with installation promised within 48 hours.
• Real time: Anomaly detection must respond within a few seconds.

🧑‍💻 The stack

• Programming Languages: Python, SQL (R is a plus)
• Data engineering: ETL/ELT pipelines, Airflow, dbt, Spark
• Cloud: AWS or Azure
• MLOps: Deployment, Standardization, and Monitoring of Models in Production
• Reporting: data visualization tools, documented APIs
• Sources: real-time machine data, OPC-UA protocol, sensors, flat files, API

🎯 Ideal Candidate Profile

• Experience: Senior-level profile, with models that have actually gone into production—not just completed notebooks.
• Expertise: statistical modeling and machine learning, anomaly detection, interpretability methods.
• Versatility: You're comfortable working across the entire workflow, from ingestion to production. In this role, there's no one else to do it.
• Data architecture: You know how to design data storage and flow, and establish standards.
• Business acumen: You can explain how much your business model generates, whether in euros or in terms of return on investment.
• Bonus: Experience in an industrial setting, working with sensor or production data.
• Languages: Fluent in French, business-level English.

🏢 The Work Environment

• Reports directly to the president.
• Executive status, daily flat rate.
• Position based in Paris, with a hybrid work arrangement: on-site and remote work.
• Close collaboration with the technical team and with the client-side field teams.

🛑 This job isn't for you if...

• You want to focus on modeling and leave the pipelines and deployment to someone else.
• You need a data team and a platform that's already in place.
• Accurate, certified data is essential for you to do your job well.
• The title "Head of Data" must appear on the contract at the time of signing.

✅ This job is for you if...

• You want to see your designs being used every day by people on the shop floor, not tucked away in a report.
• Building a data function from scratch motivates you more than tweaking it here and there.
• You enjoy problems where the data is messy, noisy, and unlabeled—because that's where the real work begins.
• You want to be the first data profile for a product that already has industrial customers.

📌 Prerequisites

• Fluent in French: The team, management, and some of the clients work in French.
• Valid authorization to work in France, with no action required on the part of the employer.
• Based in the Île-de-France region: This is a hybrid position requiring regular on-site work in Paris and governed by a French employment contract.

🪜 The Hiring Process

2 to 3 steps, including a meeting with the president. Allow two to three weeks.