You need these six building blocks to start AI in your organization successfully (II)

In the previous chapter, we discussed Business Cases and Data as the first two cornerstones of AI adoption.

Here we are with another two - Technology and Talent Pool.

Have a good read!

Key Takeaways:

1. The technology serves the use case, not the other way around. Learn the novelties, but don't forget about simpler, proven methods.
2. AI development is a teamwork. Quite complex, to be honest. The baseline is always within the organization. Pair your employees with reliable AI professionals to deliver the results you seek.

Technology

The dynamic nature of Big Data, Analytics, and AI technologies can create uncertainty.

Just look at how many GPT-like systems you see pop up every day.

So, how do you deal with it?

First, you don't need to know all the algorithms, models, and virtual machine properties.

What you need to know (or, let's say, find out) is how the system will work.

The bigger picture.

It is called a 'functional reference architecture.' Basically, it's a decomposition of the system—the logic of its operations.

You only need to design the data flow based on your use case.

Treat it as a roadmap for the necessary data collection, storage, management, and processing tools.

For example, if you're building a simple computer vision system, you'll need the following:

• a camera to gather images
• a server to store the images
• an ML model to process images
• a device with computing power that will run the model
• any outcome, be it an alert, user interface to see the model's predictions, etc.

That's a good start. Now imagine going deeper, as the more details you know, the more useful this document will be.

Let's say the sytem you want to build is a Quality Control station in a manufacturing plant.

• The camera will be a high-speed camera taking monochrome, high-resolution pictures.
• Images will be processed locally on a PC set up on the station.
• The model will actually be a chain of models. The first one will choose an area of interest. The second one will look for defects in the area. The third one will determine whether it's a critical defect.
• If a defect is critical, the QC operator will be alerted and will decide whether to stop production, remove the defective part, or let the production go.

Then, the project team will use such a map to design the system's final architecture.

Technological choices are secondary to the big picture.

Talent Pool Access

There's an ever-growing need for skilled professionals.

Nothing new here.

What matters, though, is how tWhat'sthe roles are in AI and ML development.

There are data scientists that can extract meaningful information from raw data. They also that's ML models to fit the kThere'spredictions you seek.

There are data engineers that perform operations on data to make it as useful as possible.

ML engineers make models operable. They build the ecosystem, design models, and improve them.

Software engineers integrate trained ML models with the business application.

You also have data stewards, solution architects, and analytics translators. And a whole lot of other roles.

Now, the questions that matter to you:

1. How big of a team does my project need?
2. What's the perfect team composition?

Big enough to do the job but small enough to be nimble. Consisting of as many roles as needed.

But that's not an answer, is it?

There's no silver bullet here. Each project has a unique scale, complexity, time to market, etc.

But, as you can always add new members, try to set the initial number.

Here's what you have to take into account:

Project Scope and Complexity

Larger, more complex projects require more personnel. This is due to the sheer number and size of aspects to manage. Data collection, model development, and integration. Ecosystem support. Someone has to do all this.

Project Phase

The early stages might need more exploratory data analysis and model prototypes. Later stages might require more software engineers for deployment and maintenance.

Expertise and Skills

A balanced team should include members with a mix of skills. They need to deal with all aspects of developing and building software around a model.

Budget and Resources

The budget available can limit the size of the team. More resources can allow for a larger team with more specialized roles. Nothing fancy here.

Timeline and Deadlines

Tighter deadlines might require a larger team to work so that it can be delivered on time.

Industry and Application Domain

Specific industries or applications require specialized knowledge or experience. Adding the domain expert, even part-time, is always a good idea.

Research vs. Application Focus

Teams focused on research and development of new ML methods can use more academic researchers. Application-focused teams prioritize engineers and developers.

This all has to be taken into account. Luckily, most AI vendors will help with the team composition.

And you can always scale the team up or down.

In the next chapter, we will cover Processes and Culture as the last elements that you need to take care of. Stay tuned!

For more ML and AI insights, subscribe or follow Sparkbit on LinkedIn.

If you're looking for a team to run your AI project end-to-end, DM us or write to [email protected]

Author: Kornel Kania , AI Delivery Consultant at Sparkbit