Project Management

Defining the goal of the AI project: The first step is to clearly define the goal we want to achieve by implementing an AI-based solution. The goal may be, for example:

• improving the efficiency of business processes,
• optimizing decisions,
• automating tasks,
• or developing new products or services.

Automating tasks is a great goal for an AI project. Do you know what task automation is? Let's check it out by watching the video below.

Using AI in your automations. Unbelievable!

Data analysis: A key element of an AI project is proper data preparation and analysis. At this stage, available data resources should be examined and their quality, availability and consistency assessed. It is

al-networks-584ababf69de

Choosing an AI algorithm and model: Based on the project goals and data analysis, you should choose the appropriate algorithm and AI model that will best suit your needs. There are many different AI techniques and models, such as:

✔ neural networks ✔ decision trees, ✔ machine learning algorithms.

d-7a7d-0010-87a3-c30de2ffd8ff

AI model training: In this stage, the AI model is trained based on available data. The model is trained to learn to recognize patterns and make appropriate predictions. It is also important to validate the model on test data to assess its effectiveness.

Let's look at training an AI model How do you train artificial intelligence:

What is AI training?

✔ neural networks ✔ decision trees, ✔ machine learning algorithms.

When you train AI, you’re teaching it to properly interpret data and learn from it in order to perform a task with accuracy. Just like with humans, this takes time and patience (just consider all of those worksheets you had to complete when learning your multiplication tables back in grade school). Only by training AI to correctly perceive information and make accurate decisions based on the information provided, can you ensure your AI will perform the way it’s intended.

You can read and watch more here:

You can read more about how to process data here bb3/TELUSInternational_DataAnnotation_eBook.pdf

After completing the model training, implementation in a real application or system takes place. It is important to provide appropriate infrastructure, such as servers, and integration with existing systems.

Let's read and see how this process has been implemented in medicine, including COVID19.

At this point, it is worth recalling the article about the use of Artificial Intelligence in medicine in a number of diagnostic processes. Due to the COVID-19 pandemic, AI is used effectively to diagnose infections based on MRI, but artificial intelligence is also known to be used in the diagnosis of scoliosis, colorectal cancer, retinal opacity and a number of others. In the first phase of AI development, reference data is collected in the resources of the hardware platform.

But how does such artificial intelligence work?

The principles of AI operation are simply illustrated in the diagram below presenting the principle of operation of the OpenVINO platform popularized by Intel and compatible with iEi hardware.

The figure below shows how images are identified and interpreted, resulting in the recognition of people in a video image.

To create an expert system, a layer supporting learning (Trained Model) is necessary, then a layer optimizing the training sequence (Model Optimizer) and converting data into dependencies in the expert system, and a layer in which the final AI code is executed (Inference Engine). For the individual layers in which the AI expert system is created, iEi offers dedicated equipment, which is particularly clear in the example of using AI to diagnose macular degeneration in the eye.

In the drawing we see three layers of the solution:

• learning support layer (left), a layer that converts
• optimizes the training sequence into dependencies in the expert system (in the middle),
• hardware layer on which the final AI code is executed (right). In this case, the working environment was based on Microsoft Azure. more here

Testing and optimization: After implementing the AI model, it is necessary to conduct tests and evaluate the results. At this stage you can also make corrections and optimize the model to obtain better results.

Implementation and monitoring: After testing, the AI model is ready for larger scale implementation. It is important to constantly monitor its operation and evaluate the results to ensure proper functioning.

It is also important that the planning and implementation of AI projects take into account appropriate regulations and ethical principles. You should also anticipate the need to update and evolve the AI model as changes and new data are introduced.

Let’s check your knowledge about Project Planning and Execution? (True or False)

What have you learned about Planning and Executing AI Projects?

• Can you define the goals of an AI project ?
• A key element of an AI project is appropriate data preparation and analysis. Do you know how to extract important features of a task and process this data appropriately ?
• Which AI model and algorithm should I choose ?
• Training, learning, pattern recognition and validation are important elements of AI effectiveness
• To properly implement an AI model, you also need to know the infrastructure (servers) and systems in which it will be implemented
• Testing, optimization and monitoring will ensure the proper functioning of the AI model.

In today's fast-paced business environment, organizations across industries are continually seeking ways to enhance productivity, accelerate project delivery, and foster collaboration among teams. Agile Scrum methodology has emerged as a powerful framework that allows teams to adapt and deliver high-quality results in an iterative and collaborative manner. Now, with the advancements in artificial intelligence (AI), teams can take their Agile Scrum practices to new heights, leveraging AI tools to further streamline processes,

• Can you define the goals of an AI project ?
• A key element of an AI project is appropriate data preparation and analysis. Do you know how to extract important features of a task and process this data appropriately ?
• Which AI model and algorithm should I choose ?
• Training, learning, pattern recognition and validation are important elements of AI effectiveness
• To properly implement an AI model, you also need to know the infrastructure (servers) and systems in which it will be implemented
• Testing, optimization and monitoring will ensure the proper functioning of the AI model. automate tasks, and optimize outcomes. In this lesson we will look at the Agile Scrum methodology.

Agile, Scrum in Data Science / AI Development / Machine Learning Scrum is often understood in the context of software development. There is a well-regulated craft, well-accepted good practices and a well-understood process. Outside the software development trend, the data science movement has evolved from the world of mathematicians and statisticians.

Because these worlds have different origins, I often hear that IT development tools do not fit into the development of artificial intelligence models. I would like to dispel popular myths about the use of agile methods when creating models.

Challenges of creating AI models.

In addition to selecting the parameters of the model's architecture, data for learning is a great challenge. Apparently it used to be said that "what is input is what is output", if we have low quality data for learning, our gnome will not be a very smart gnome - rather one who simply "passed" the junior high school exam. When an engineer receives data from a client, more than half of his work is analyzing this data, identifying ambiguities and correcting them. Imagine that a client would like to have such a clever gnome who would guess based on the invoice whether it will be paid on time or not. Such a client provides archived invoices and data on whether he has paid or not. A problem may arise when, during the pandemic year of 2019, the client changed the invoice program and the new invoices have an additional "disinfection costs" field. Our gnome may notice that invoices with such a field are paid very late (because during the pandemic, contractors have liquidity problems) and then he will pay more attention to this field instead of, for example, the field with the recipient of the invoice or the type of service sold.

The goal of the AI engineer is to thoroughly check the training data (for training the gnome), clarify dependencies with the client, understand anomalies, and often require a deep understanding of the client's business processes.

Development in AI Therefore, a wise engineer receives a package of data from the client, analyzes it, teaches a simple gnome and checks the learning results, carefully taking notes in a notebook with a space pencil. It then corrects the input, changes the gnome's parameters, trains it longer, and checks the results again. There are so many iterations until the gnome is as smart as the client needs it to be.

.. sounds like SCRUM?

Look what we have:

• complex problem in a chaotic environment,
• pivots are difficult to predict,
• many iterations,
• set goals, measurable results,
• a team of smart people in flannel shirts,
• focus on achieving the result. Sounds like the perfect problem to implement agile methods, right?

In the second block, we learned that in the context of artificial intelligence, the use of Agile Scrum involves integrating the cooperation of programmers, data scientists, analysts and experts in the field of AI and has a huge impact on its development.