Data science—no longer just for data scientists

Pyramid simplifies the ML lifecycle for both citizen data scientists and expert data scientists. In Data Flow—an intuitive visual interface that represents the data pipeline—users can drag and drop powerful AI-driven data transformations and algorithms directly in the flow or using scripting. They can leverage deep Python and R integrations to perform feature engineering, build and apply ML models, and orchestrate data pipelines. The low code /no code, automated approach helps data scientist accelerate through tedious tasks in a data science project.

Data scientists spend 80 percent of their time finding, cleansing, exploring, and organizing data. Pyramid’s drag-and-drop modeling tool reduces the model development lifecycle with AI-driven tools for data cleansing, wrangling, and transformations. Users can explore the shape of the input data with advanced visualizations and robust statistics for exploratory data analysis to identify dataset characteristics. Use natural language to generate SQL, DAX, and MDX code automatically using our ChatGPT integration.

Pyramid provides out-of-the-box, pre-built algorithms that can easily be used by business analysts and citizen data scientists. They can apply these algorithms directly within the intuitive, drag-and-drop flow interface to perform sophisticated machine-learning tasks without any coding. Algorithms can be downloaded from the Pyramid Marketplace, selected from Pyramid’s content management system, or simply pasted into the script window.

Solve is an engine that can generate prescriptive optimization solutions based on decision models designed through a no-code/low-code spreadsheet-like Tabulate interface. Business users can build models with complex logic waterfalls, conditions, and decision points, using the entire suite of spreadsheet formulas. Solve delivers true decision intelligence on live data—answering both “What should I do” questions and “What if” simulations that can be embedded into dashboards and presentations.

Pyramid has made Python (and R) a first-class citizen in its architecture. The integration enables data scientists to easily train models and deploy them into production. A marketplace with free, reusable Python source code provides non-technical analysts with a library of predefined functions. And existing Python scripts can be further customized to suit more specific business requirements.

Feature Engineering in Pyramid enables data scientists and business analysts to transform raw data into features suitable for ML models. Apply transformations or use math functions to create new features. Access features from other platforms like Databricks that provide API access to features. And share engineered features in Pyramid via rest APIs for use in other applications as well.

Build and train models using learn and predict scripts or regular scripts. Users can run the learn function against training data (input) and return the ML model (output). Users have access to an extensive library of scripts, accessible through simple point-and-click prompts, including Weka, MLIB, and TensorFlow—as well as DL4J for deep learning. Pyramid features pre-built scripts to build and train models. Users can train models using various approaches to achieve faster results and higher accuracy, or Models can be saved and re-used against other data sets and use cases. Apply your saved model to new data for testing or scoring by simply connecting it to the new data flow.

Models created on Pyramid, along with their required data pipeline, can be saved, scheduled and executed on Pyramid directly. The output of the models can be monitored on the platform to provide information on model performance over time and identify model degradation and drift. Alerts can be set to inform users when the model performance changes so the model can be retrained.

The Pyramid Marketplace contains pre-built ML scripts you can apply for data cleansing (outlier analysis, delete empty rows, etc.), data prep and transformations tasks, statistical analysis, and classic ML regression and classification tasks such as forecasting and clustering. It features pre-built scripts for classic ML modeling tasks like sentiment analysis, market basket analysis, geospatial modeling, and more. Marketplace also includes scripts for specialized use cases like modeling stock performance, future insurance claims, infant health, and more.