Radar
Technologies
| Name | Category | Recommendation | Description | Link |
|---|---|---|---|---|
| Scikit-Learn | Machine Learning | ADOPT | General ML library | scikit-learn.org ↗ |
| pytorch | Machine Learning | ADOPT | Deep learning library | pytorch.org ↗ |
| xgboost | Machine Learning | ADOPT | Special implementation of gradient-boosted trees | xgboost.readthedocs.io ↗ |
| lightgbm | Machine Learning | ADOPT | Special implementation of gradient-boosted trees | lightgbm.readthedocs.io ↗ |
| mlflow | ML Ops | ADOPT | Experiment tracking and model registry | mlflow.org ↗ |
| explainerdashboards | ML Ops | ADOPT | Interactive dashboards for model evaluation; computes model-agnostic KPIs (AUC, F1, etc.) and charts (ROC, PR, etc.) for classification and regression, as well as permutation feature importances and SHAP values. | explainerdashboard.readthedocs.io ↗ |
| shap | ML Ops | ADOPT | Shapley values for ML explainability | github.com ↗ |
| Google ADK | Gen AI | ADOPT | Develop your own agents with this | google.github.io ↗ |
| Google Gemini | Gen AI | ADOPT | Google's LLM | gemini.google.com ↗ |
| VertexAI | Gen AI | ADOPT | GCP AI Platform | cloud.google.com ↗ |
| Spark | Data Wrangling | ADOPT | Framework for distributed data processing, including ML models, e.g. via Google DataProc or Snowpark | spark.apache.org ↗ |
| Snowflake | Data Wrangling | ADOPT | Cloud DWH | snowflake.com ↗ |
| polars | Data Wrangling | ADOPT | DataFrame library written in Rust | pola.rs ↗ |
| pandas | Data Wrangling | ADOPT | DataFrame library | pandas.pydata.org ↗ |
| dbt | Data Wrangling | ADOPT | build SQL models easily | getdbt.com ↗ |
| streamlit | Dashboarding | ADOPT | Dashboarding library | streamlit.io ↗ |
| Microstrategy | Dashboarding | ADOPT | Graphical tool for dashboards, etc. | strategy.com ↗ |
| PowerBI | Dashboarding | ADOPT | Graphical tool for dashboards, etc. | microsoft.com ↗ |
| Adobe Analytics | Dashboarding | ADOPT | Graphical tool for interaction data from our websites and apps | business.adobe.com ↗ |
| plotly | Plotting | ADOPT | Quickly build interactive charts | plotly.com ↗ |
| matplotlib | Plotting | ADOPT | Old-fashioned but powerful plotting library | matplotlib.org ↗ |
| seaborn | Plotting | ADOPT | Matplotlib wrapper for various statistical plots | seaborn.pydata.org ↗ |
| uv | Python Tools | ADOPT | Python package manager written in Rust by Astral | docs.astral.sh ↗ |
| poetry | Python Tools | ADOPT | Python package manager | python-poetry.org ↗ |
| ruff | Python Tools | ADOPT | Linter and formatter | docs.astral.sh ↗ |
| black | Python Tools | ADOPT | Formatter | github.com ↗ |
| mypy | Python Tools | ADOPT | Static typing | mypy-lang.org ↗ |
| Airflow | Orchestration | ADOPT | Industry standard for orchestration | airflow.apache.org ↗ |
| Dagster | Orchestration | ADOPT | Orchestration tool | dagster.io ↗ |
| pydantic | Data Validation | ADOPT | Data validation library | docs.pydantic.dev ↗ |
| statsmodels | Other | ADOPT | Statistics library | statsmodels.org ↗ |
| numpy | Other | ADOPT | Numerical computing tools in Python | numpy.org ↗ |
| scipy | Other | ADOPT | Versatile library for scientific computing in Python (e.g., optimization, integration, differential equations, statistics, etc.) | scipy.org ↗ |
| TabPFN | Machine Learning | ASSESS | Pretrained transformer specialized for small tabular datasets | github.com ↗ |
| feast | ML Ops | ASSESS | Feature Store | feast.dev ↗ |
| ollama | Gen AI | ASSESS | Run open LLMs locally | ollama.com ↗ |
| litellm | Gen AI | ASSESS | LLM Gateways for Model Access | litellm.ai ↗ |
| mlflow | Gen AI | ASSESS | Model registry and many other things; lately heavy development for GenAI use cases | mlflow.org ↗ |
| vllm | Gen AI | ASSESS | Serving local LLMs | vllm.ai ↗ |
| DAIP Apigee | Gen AI | ASSESS | DAIP LLM abstraction layer; call different models via DAIP | |
| Dask | Data Wrangling | ASSESS | Distributed Python data wrangling with familiar packages | |
| narwhals | Data Wrangling | ASSESS | Library-agnostic DataFrame operations; useful for cross-cutting functions that should work with pandas, polars, Spark, or duckDB. | github.com ↗ |
| ray | Data Wrangling | ASSESS | Scale infrastructure for analytical workloads | ray.io ↗ |
| Shiny | Dashboarding | ASSESS | Dashboarding library | shiny.posit.co ↗ |
| Custom Solution | Dashboarding | ASSESS | Custom solution created with agentic engineering, e.g., in React | |
| altair | Plotting | ASSESS | Grammar of Graphics plots in Python | altair-viz.github.io ↗ |
| pygwalker | Plotting | ASSESS | Interactive data analysis of DataFrames; looks a bit like Tableau | github.com ↗ |
| ty | Python Tools | ASSESS | Static typing written in Rust by Astral | github.com ↗ |
| dataframely | Data Validation | ASSESS | Lightweight library for validating polars DataFrames | dataframely.readthedocs.io ↗ |
| patito | Data Validation | ASSESS | Validates DataFrame objects based on Pydantic | github.com ↗ |
| tensorflow | Machine Learning | HOLD | Deep learning library by Google | tensorflow.org ↗ |
| jax | Machine Learning | HOLD | Array-oriented library for deep learning research | docs.jax.dev ↗ |
| prophet | Machine Learning | HOLD | Simple time series modeling by Meta | facebook.github.io ↗ |
| kubeflow | ML Ops | HOLD | ML pipelines on Kubernetes (e.g., also VertexAI) | kubeflow.org ↗ |
| metaflow | ML Ops | HOLD | Library for the complete ML lifecycle by Netflix | metaflow.org ↗ |
| deepchecks | ML Ops | HOLD | Model monitoring library, e.g., for feature drift | github.com ↗ |
| eli5 | ML Ops | HOLD | Model explainability | eli5.readthedocs.io ↗ |
| Snowflake ML | ML Ops | HOLD | Set of tools and capabilities from Snowflake for data scientists. Includes:Model RegistryFeature StoreExperiment TrackingModel Explainability | docs.snowflake.com ↗ |
| langfuse | Gen AI | HOLD | Open SourceLLM Engineering Platform | langfuse.com ↗ |
| langgraph | Gen AI | HOLD | lang* agent system | langchain.com ↗ |
| pocketflow | Gen AI | HOLD | open source agent framework | github.com ↗ |
| LM Studio | Gen AI | HOLD | Run AI models, locally and privately. | lmstudio.ai ↗ |
| langchain | Gen AI | HOLD | lang* abstraction layer | langchain.com ↗ |
| Apache Pinot | Data Wrangling | HOLD | Real time distributed OLAP data store, purpose-built to provide ultra low-latency analytics at extremely high throughput. | pinot.apache.org ↗ |
| Apache Beam | Data Wrangling | HOLD | Highly abstracted stream and batch processing. For example, it is behind Google Dataflow. | beam.apache.org ↗ |
| dash | Dashboarding | HOLD | Dashboarding library | dash.plotly.com ↗ |
| reflex | Dashboarding | HOLD | Dashboarding library | github.com ↗ |
| Tableau | Dashboarding | HOLD | Graphical tool for dashboards, etc. | tableau.com ↗ |
| Excel | Dashboarding | HOLD | ||
| Snowflake Dashboards | Dashboarding | HOLD | Snowflake-internal solution for dashboards | docs.snowflake.com ↗ |
| bokeh | Plotting | HOLD | Interactive plots | bokeh.org ↗ |
| Skypilot | Orchestration | HOLD | Orchestration of ML/AI workloads | github.com ↗ |
| Prefect | Orchestration | HOLD | Orchestration tool | prefect.io ↗ |
| Google Workflows | Orchestration | HOLD | Simple orchestrator in GCP | cloud.google.com ↗ |
| cupynumeric | Other | HOLD | NumPy on NVIDIA GPUs | docs.nvidia.com ↗ |
| PyMC | Machine Learning | TRIAL | Library for creating Bayesian models (keyword: MCMC) | pymc.io ↗ |
| VertexAI Model Registry | ML Ops | TRIAL | Model Registry within VertexAI | |
| langsmith | Gen AI | TRIAL | Closed SourceLLM Engineering Platform | eu.smith.langchain.com ↗ |
| Custom Evaluation | Gen AI | TRIAL | Custom evaluation - whenever LangSmith or MLflow are not enough | |
| Snowflake Cortex | Gen AI | TRIAL | Snowflake’s LLM | snowflake.com ↗ |
| duckDB | Data Wrangling | TRIAL | Fast in-memory database system. Read data and process it extremely quickly with SQL. | duckdb.org ↗ |
| ibis | Data Wrangling | TRIAL | DataFrame library for different backends | ibis-project.org ↗ |
| VertexAI Pipelines | Orchestration | TRIAL | Managed Kubeflow Pipelines in VertexAI | docs.cloud.google.com ↗ |
| copier | Other | TRIAL | Management of code project templates; changes to the template can be pushed to repositories that use the template. | copier.readthedocs.io ↗ |
| H20 | Machine Learning | UNKNOWN | General ML library | docs.h2o.ai ↗ |
| Nixtla Statsforecast | Machine Learning | UNKNOWN | Time series forecasting with statistical and econometric models | nixtlaverse.nixtla.io ↗ |
| Catboost | Machine Learning | UNKNOWN | Gradient-boosted trees specialized for categorical features | catboost.ai ↗ |
| AutoGluon | Machine Learning | UNKNOWN | AutoML package | auto.gluon.ai ↗ |
| trackio | ML Ops | UNKNOWN | Lightweight experiment tracking by Hugging Face | github.com ↗ |
| skore | ML Ops | UNKNOWN | Model-agnostic automatic evaluation including cross-validation of models | github.com ↗ |
| featuretools | ML Ops | UNKNOWN | Automatic feature engineering | github.com ↗ |
| faker | ML Ops | UNKNOWN | Generation of test data | fakerjs.dev ↗ |
| shap-iq | ML Ops | UNKNOWN | shapiq: Shapley values and Shapley interactions for ML explainability (extension of shap) | github.com ↗ |
| argilla | ML Ops | UNKNOWN | Collaboration tool for data scientists and domain experts to label data and evaluate models | argilla.io ↗ |
| Weights and Biases | ML Ops | UNKNOWN | Commercial experiment tracking and model explainability, primarily for deep learning models | wandb.ai ↗ |
| Captum | ML Ops | UNKNOWN | Model explainability for PyTorch by Meta | captum.ai ↗ |
| uqlm | Gen AI | UNKNOWN | Detect hallucinations | github.com ↗ |
| Ai Studio | Gen AI | UNKNOWN | LLM Playground by Google | aistudio.google.com ↗ |
| python-genai | Gen AI | UNKNOWN | Python Lib to use Google LLMs / LMMs | github.com ↗ |
| Meerschaum | Data Wrangling | UNKNOWN | Time Series Data processing | github.com ↗ |
| nicegui | Dashboarding | UNKNOWN | Dashboarding library | nicegui.io ↗ |
| taipy | Dashboarding | UNKNOWN | Dashboarding library | taipy.io ↗ |
| zuban | Python Tools | UNKNOWN | Static typing written in Rust | github.com ↗ |
| pyright | Python Tools | UNKNOWN | Type checker | github.com ↗ |
| basedpyright | Python Tools | UNKNOWN | Open-source fork of pyright | github.com ↗ |
| pandera | Data Validation | UNKNOWN | Validates DataFrame objects | pandera.readthedocs.io ↗ |
| StringZilla | Other | UNKNOWN | Accelerates string operations (by up to 10x) in C, C++, Python, Rust, Swift & Go | github.com ↗ |
Analytics Radar Target Vision
Our Analytics Radar is a practical and collaborative guide that helps colleagues and teams navigate new technologies, libraries, and software solutions. It provides a structured expert assessment of tools and applications that may be relevant for us, including classification into labels such as “Adopt” or “Assess” (more on this in the next section below). The focus is not on universal recommendations or completeness, but on concrete experience from our day-to-day work.
We invite everyone to contribute their assessments and practical experience: feedback is explicitly welcome and essential for the further development of the Radar. The process is deliberately open. Every few weeks, we focus on a specific topic area and regularly revisit all topics to reassess and further develop them. Our focus is on open-source solutions or technologies that have already been purchased, such as Snowflake or GCP. Purchasing advice or bans are not part of the Radar.
The goal is a living, collectively supported tool that provides orientation without being restrictive. A Radar that grows with us and continuously evolves.
Categories
We use the same categories as Thoughtworks or Zalando.
Label | Description |
|---|---|
| ADOPT | Clear recommendation for production projects. |
| TRIAL | Initial successful projects have been implemented; some risk remains. |
| ASSESS | Recommended for early adopters, as the technology definitely has potential. |
| HOLD | Not recommended for new projects. |
| UNKNOWN | Not tested yet. |