Connect and query any data in its original format and in real-time without any preparation. RAW code generates source specific adaptors for each source optimized for high performance data retrieval. RAW automatically detects the format, the schema and the encoding to make it transparent to the user. RAW supports both simple and complex data including CSV, Excel, Word, Amazon S3, machine logs, streams, JSON, HJSON, XML, AVRO, HDFS, HDF5, Parquet, NoSQL relational data bases, columnar databases, multi-dimensional arrays, APIs and more.

RAW's extended version of SQL support all functions necessary to create workable data sets. Join, clean and transform data with different file formats all at the same time. Applying data cleansing algorithms to correct and improve the quality of your data. No need to create tables or perform heavy ETL processes. RAW supports very complex queries such arbitrarily nested queries allowing you to mix together incompatible data.

Use your favourite data science tools to turn your datasets into Smart Datasets. As a data scientist you can quickly bring together data, or use the queries created by your colleagues to run your experiments. Add new predictive data points directly from e.g. Python notebooks using Scikit Learn. Save the enhanced data as a Smart Dataset and make it automatically available in the RAW Virtual Lake. If your model works well, save it and share it with your colleagues. You can even embed your model directly in a RAW query and run it on the Edge.

Datasets are automatically virtualized in RAW either in near real-time, or through high performance caches. RAW's caching engine takes into consideration the structure and format of the datasets and optimizes the caching accordingly. A simple dataset based on CSV files will be cached different than a data set based on multidimensional arrays. RAW updates its caches as the datasets grow without having to refresh the entire cache, saving time and maintaining high response times. RAW allows the users to name and describes its datasets, thus providing a simple, yet effective to build on a repository of datasets that can then be shared with the appropriate user access rights across the organisation.