Renumics GmbH

Renumics Spotlight brings data-centric AI to your stack. Enrich your data, find problems, and fix them to train robust ML models.

How It Works


import pandas as pd
from renumics.backstage.dataset 
                    import Dataset

df = pd.read_csv("some.csv")
with Dataset("myDataset.h5", 'w') 
    as dataset:
        dataset.
            from_dataframe(df, index=False)

Enrich and Import Your Data

Import you data and enrich it with embeddings from pre-trained models with our easy-to-use Python API. Many engineering data types are supported out of the box.

Understand the Problem Space

Interactively explore your data and adapt the visualizations on the fly. Identify and annotate segments, outliers and correlations in your data.

Develop Your Data Strategy

Identify relevant populations to plan data acquisition and labeling campaigns. Assess the possibility for data augmentation and synthetic data.


import pandas as pd
from renumics.backstage.dataset 
                    import Dataset

df = pd.read_csv("some.csv")
with Dataset("myDataset.h5", 'w') 
    as dataset:
        dataset.
            from_dataframe(df, index=False)

Import Data and Model Results

Build your dataset with our easy-to use Python API. Use the embedding of your model as a similarity measure and optionally add more information and metadata.

Boost Dataset to Improve Model

Use your team's domain knowledge to understand segments, edge cases and outliers. Choose which data should be annotated or acquired next.

Re-Train and Iterate

Iterate the data boosting process until performance goals are met. Build user trust by clearly understanding the limits of the model.

Frequently Asked Questions

How do I know that Renumics Spotlight is right for me?

Who can benefit from Renumics Spotlight?

Teams who develop ML models and solutions that address engineering and manufacturing processes. Teams that have both experienced data scientists with Python skills and non-coding domain experts tend to benefit the most from Spotlight.

Is there a requirement for using Spotlight?

At least one team member must have basic Python skills to load the data into Spotlight via our API.

I’m building my own data curation tools. Why should I use Spotlight?

If you've started to build you own tooling for data exploration and curation, you have experienced how important this topic is. We have spent a lot of effort building a high-performing solution that is intuitive to use and gets not only the big picture, but also the little details right. There are two main advantages for you: Firstly, you save the time to build your own tooling and you can start right away. Secondly, Spotlight allows for a true interactive exploration even for non-coders that cannot be achieved by a dashboard. In a meeting with all stakeholders, this difference can be night and day (i.e. weeks of project time). Please contact us for a demo to experience the difference yourself.

Do you provide ML models for my use case?

We believe that a modular stack for ML is the best approach to data-driven engineering. We focus on providing the best data curation tool possible that seamlessly integrates into existing ML and engineering toolchains. Following this philosophy, we do not provide specific modeling tools. However, we feel that unsupervised and explainable ML is a crucial aspect of great data curation workflows. Thus, we explore these topics within publicly funded research projects and are helping to educate our partners and customers through blog material, tutorials and workshops.

Does Spotlight allow to deploy ML models?

We see Spotlight as a really good data curation tool that integrates well with other solutions. We therefore do no provide specialized deployment functionality but encourage our users to rely on existing MLOps tools that suits their application and IT environment. If you need guidance to navigate the MLOps tooling landscape, please do not hesitate to contact us.

What data types and formats do you support?

A big strength of Spotlight is the ability to handle a wide variety of data types that arise in engineering and manufacturing. This includes general types such as geometries, time sequences or images. We also support application-specific data like deformations, pressure fields or acoustic sequences. In terms of formats we support standard interfaces (e.g. HDF5), but we also integrate with libraries that allow to process more specialized data types. If you have a special format in mind, please do not hesitate to contact us.