Engineering the AI applications (Part 1) - a balancing act between data and knowledge representation

How we started learning AI development?

In this short post, I would like to share a few lessons that we have learnt while developing the AI applications. Started collecting datasets, simulating data. I started feeling I was disconnected; that we were missing something. We are very clear of what we do with the data, but we could not relate to the business context.

I had an interview with one of the business owners who has no technical background and bias. He said "I run the business for last 13 years and I learned a lot of insights. Now why I should rely on your data?".

Back to the basics:

We realized that we were overwhelmed by the data all the time and ignored the fundamental ingredient of AI system - the Knowledge Representation.

We have reworked a process to bring both the data and knowledge representation in our Platform in a simple way.

One size doesn't fit for all:

This is the most important decision we made. Every AI application (agent) is purposely built to assist the business process.

Avoid building a universal AI agent!

Essential Process steps we learned:

• Establish a few benchmark parameters from the existing datasets. Get them reviewed by the customers or domain experts.
• Prepare an unambiguous questionnaire to represent the specific knowledge from the customers.
• Prepare a dynamic state-space representation.
• Identify required mathematical models (Operators) that require to apply at each state.
• Software + hardware development.

Take away:

• We could improve in scoping the AI applications and focus on solutions that address customer problems.
• Avoid building a universal AI agent!
• Improve on compute requirements to run the AI models (rules).
• Work closely with the customers.

Vidcentum Technologies is open to offer AI, IT/OT related services.