AI Prototype Building - 10 Day Implementation


Start to build a prototype of your AI idea / solution to profoundly analyze the possibilities and the impact on business.

Dataroots provides an AI prototype & deployment environment on the Azure cloud platform. We foresee easy access to data sources and pre-defined storage options to efficiently prepare the data for further processing such as AI modelling. Using AI developer tools, models can be built, tested and validated and exposed via an API or dashboard. It will allow for a fast assessment of the possibilities and feasibility of building AI solutions on top of your data. From a business perspective we will, together with the client, assess how the Azure Cloud Platform contributes to automating & productizing AI solutions in order to generate real-life impact (ROI).

The deliverable is a tangible end-to-end AI prototype built on the Azure Cloud Platform, using the state-of-the-art Azure services with an exposable result towards the end user.

The agile approach we take in our prototype building is the following. (In all phases the client will be directly involved, through intense deliberation and co-creation).

  1. Business understanding (±10%) Determine the scope and the objective of the project together with the client. Why does the project bring value to the organisation? At the end of this phase we have a clear scope and goal

  2. Data understanding (±30%) Collect and understand all the necessary data through data exploration and visualization. The goal is to have a clear understanding of the data available.

  3. Prepare the data (±30%) Clean the data, create descriptive insights, feature engineering and prepare the data for model ingestion. This leaves us with cleaned data and a descriptive report for the business (using the visualisation of the previous step as well).

  4. Modelling (±20%) Test different models that transform raw data into actionable insights. At the end of this phase a modelling report will be delivered.

  5. Evaluate and deliver (±10%) Evaluate and select the best performing model. We deliver the code base of the model and a visual tool to interpret the results