Data Science is a new field. While we seen its transformational power in many industries, we also understand that some organizations might have reservations related to adopting new technologies. That risk aversion should not be a reason to lag on the market behind other competitors.
A Proof-of-Concept is a step-by-step approach that starts with a simple Data Science project. The use-case is intended to be conducted in a short timeframe and with a small budget. The goal of the PoC is to obtain tangible insight regarding the feasibility of the project for the organization and derive measurable results even with a small dataset and toy model.
As such, the Proof-of-Concept might as well be the bridge towards becoming a AI-driven enterprise.
A PoC test campaign starts once an appropriate Data Science project is selected and the neccesary data is secured.
The data used in the PoC campaign is much more narrow and downsized compared to a full-fledged project. In particular, our Data Scientists will only ask for the minimum valiable dataset neccessary to build a model that is precise enough to show-case the value of the technology.
In addition, other components of a Data Science project such as Feature Engineering and Data Vizualization will also be limited. Typically, no dashboards and applications are provided.
The PoC test campaign does not utilize our data infrastructure. In general, our Data Scientists will use powerful workstations, but not advanced systems such as Hadoop or Apache Spark given the smaller datasets and less complex models.
No two PoC campaigns are the same. Therefore, the effort and time is discussed with the client as a function of the business use-case.
Most projects can be separated into two phases: the development of the predictive model and the validation of the model through a campaign.
The development of the model on average takes between 4 and 6 weeks, depending on the complexity of the project. Our Data Scientists use proprietary tools and libraries that will speed up the process as much as possible. However, our experience has shown that this is typically the amount of time necessary to build a viable model.
The PoC campaign presents an opportunity for your organization to explore the application of a cutting-edge technology.
Through the utilization of the guidance and expertise of our team, consider the PoC as a turnkey playground to quickly and safely test and validate the value of an AI project for a fraction of the time and cost it would take to implement a full-fledged project.
More importantly, the progress obtained through the PoC can be converted towards a full-fledged project, whenever you decide to put the solution in production. Thereby, your organization would be optimizing the costs in each phase of development.
In general, the PoC result in respect of the direct output is (1) a predictive model; (2) power point presentations with a summary view of the project; (3) guidelines for validating the model in production and (4) documentation of the project.
In addition, each PoC is expected to fulfill the following goals:
- The model should be reusable at least couple of times in a particular business scenario;
- The model should provide clear and measurable feedback regarding the feasibility of the solution in production;
- The PoC should provide a straightforward way to convert and upgrade the progress in a full-fledged solution.