Full Scale AI Projects

over 40 delivered AI/ML projects

A typical AI/ML project includes three phases: proof of concept, pilot and production use. Often times POC phase is only for testing machine learning part which is naturally vital. However, it’s normally only a small part of the whole “production ready” AI solution.

We have implemented several full scale AI/ML projects, so we know the pitfalls and gimmicks in these projects.

Proof of Concept (POC)

At first it’s important to make sure that a selected business case can be solved by using machine learning. It means that we can find meaningful patterns from data and usually make good enough prediction or clustering models. And if we can’t, then there is no point to continue further. Fail fast and cheap works in this context.

In this phase we use good old CRISP-DM methodology.

All POC’s are different but a usual POC project takes about 1-2 calendar months to complete.

If you don’t have sound ideas for a POC phase, maybe you could try to find some in a AI-Workshop?

Pilot and Production use

After a successful POC project we know that machine learning works as we expected. In a pilot project we expand the scope and start delivering results for selected users or user groups.  If the pilot project is successful as well, then the next step is to expand and operationalize whole solution into production use.

It’s quite common that a time interval between a successful POC and Production use can be 1-2 years. Check out these case examples of projects we’ve delivered.

Common use cases
  • Customer analytics, predictions (demand, churn, cross/up-selling, etc.)
  • Customer profiles / segments
  • Volume, sales and demand predictions
  • Pricing
  • Predictive maintenance
  • Quality control
  • Assembly clustering
  • Anomaly detection
  • Iot related machine learning
AI/ML Tools and Technologies

Purely in machine learning context we rely mostly on Python and R languages.

Our Bilot AI-Core solution is a framework of reusable and highly modularized R/Python code. In practice, Bilot AI-Core is an accelerator which minimizes a coding work and automates several work phases. Let us know, if you are interested in.

In addition, we have implemented AI projects also with RapidMiner and Azure Machine Learning Studio.

In Azure we prefer Databricks and Machine Learning Services. Most of our solutions have been built into Azure, but AWS, Google and SAP clouds, are good options for us too.

Need help?

Want to know more? Drop us a line and we’ll get in touch.

Mika, Business Lead

mika.laukkanen@bilot.fi
+358 40 845 8432

 

Teemu, Analytics Sales Lead

teemu.lainiola@bilot.fi
+358 44 500 8031