Senior Data Scientist – Python, ML & Signal-Processing

About the company

DATATRONiQ is a German deep-tech startup focused on Industrial IoT and Edge AI. In this role, you will build and train machine learning solutions using real machine and sensor data from live production environments worldwide, ranging from mid-sized manufacturers to large enterprise groups. The work centers on anomaly detection, predictive maintenance, and real-time quality monitoring, with models that directly influence whether industrial equipment keeps running as planned or experiences unexpected downtime.

Role overview

As a Senior Data Scientist, you will own the complete data science workflow from data exploration and feature engineering through model training, validation, and deployment. Depending on the customer setup, solutions may run on edge gateways, on-premise servers, or in the cloud. The data comes from industrial control systems via OPC-UA and MQTT, so feature engineering here means working with noisy sensor signals and applying signal-processing techniques rather than just handling clean tabular data. Models are quantized, exported to ONNX, and deployed close to where production needs them, which requires careful attention to model choice, latency, and memory usage.

Technology stack

The core stack includes Python, PyTorch or scikit-learn, ONNX for edge deployment, and standard MLOps tools for version control and reproducibility.

How the team works

You will work in a small, closely coordinated team where code reviews and pair programming are part of the normal workflow. Fridays include show-and-tell sessions where the team shares useful tools and interesting discoveries from the web. You will collaborate closely with data engineers and backend developers: you build the models, they build the pipelines, and together you bring both into production.

Responsibilities

• Develop, train, and validate machine learning models for anomaly detection, predictive maintenance, and quality monitoring using real production data and industrial time-series signals.
• Design features from noisy machine and sensor data sourced through OPC-UA, MQTT, MES exports, and similar industrial inputs, including signal processing and filtering.
• Deploy models to the required runtime environment, whether an edge gateway, on-premise server, or cloud setup, including quantization, ONNX export, tuning, and field monitoring under limited CPU and RAM conditions.
• Partner with data engineers and backend developers to ensure models operate reliably inside production pipelines rather than only as notebook prototypes.
• Evaluate model success by business and operational impact, such as reducing unplanned downtime, improving throughput, and lowering defect rates, not only by F1 or AUC scores.
• Contribute actively to product roadmap discussions and technical decisions, bringing informed opinions rather than simply executing tickets.

Requirements

• A completed degree in Data Science, Computer Science, Mathematics, Physics, or a related discipline.
• At least 3 years of hands-on experience with Python, common ML frameworks such as PyTorch or scikit-learn, and deploying models in production environments.
• Practical experience with time-series analysis and signal processing, with enough understanding to recognize why a simple MLP may fail on noisy industrial signals.
• Basic MLOps knowledge, including model and data versioning, reproducible pipelines, and testing for ML code.
• Strong English communication skills, both written and spoken.
• Ability to explain and defend technical decisions within a team, including standing by a well-supported viewpoint when needed.
• Preferred extras include experience with edge deployment tools such as ONNX, TensorRT, and quantization, industrial protocols like OPC-UA and MQTT, or LLMs for chat and agentic tasks.

Working style and additional information

The role offers end-to-end ownership, from designing data capture in production through pipeline development and model inference across edge, on-premise, or cloud environments. The team decides its own architecture, tooling, and test coverage. The company also experiments early with agentic development tools such as Codex and Claude Code, adopting what proves useful. The position is mainly on-site in Stuttgart, Ulm, or Berlin, and the projects serve industrial IoT customers worldwide, from medium-sized manufacturers to DAX-listed enterprises.

Closing note

At DATATRONiQ, you are expected to shape the product, contribute ideas, and help solve challenging industrial problems with real impact in manufacturing. Interested candidates are invited to get in touch by email. Recruiting agencies and headhunters are requested not to contact the company.