Course: AI Apprentice to AI Architect - Part 3 AI Practitioner

AI Practitioner is a cross-industry advanced AI Developer position that has a growing demand in the modern world. Candidates for this role need to demonstrate proficiency in optimizing and tuning AI solutions to deliver the best possible performance in the real world. AI Practitioners require more advanced knowledge of algorithm implementations and should have a firm knowledge of latest toolsets available. In this course, you'll be introduced to the AI Practitioner role in the industry. You'll examine an AI Practitioner's skillset and responsibilities in relation to AI Developers, Data Scientists, and ML and AI Engineers. Finally, you'll learn about the scope of work for AI Practitioners, including their career opportunities and pathways.

Optimization is required for any AI model to deliver reliable outcomes in most of the use cases. AI Practitioners use their knowledge of optimization techniques to choose and apply various solutions and improve accuracy of existing models. In this course, you'll learn about advanced optimization techniques for AI Development, including multiple optimization approaches like Gradient Descent, Momentum, Adam, AdaGrad and RMSprop optimization. You'll examine how to determine the preferred optimization technique to use and the overall benefits of optimization in AI. Lastly, you'll have a chance to practice implementing optimization techniques from scratch and applying them to real AI models.

Tuning hyper parameters when developing AI solutions is essential since the same models might behave quite differently with different parameters set. AI Practitioners recognize multiple hyper parameter tuning approaches and are able to quickly determine best set of hyper parameters for particular models using AI toolbox. In this course, you'll learn advanced techniques for hyper parameter tuning for AI development. You'll examine how to recognize the hyper parameters in ML and DL models. You'll learn about multiple hyper parameter tuning approaches and when to use each approach. Finally, you'll have a chance to tune hyper parameters for a real AI project using multiple techniques.

Microsoft Cognitive Toolkit provides powerful machine learning and deep learning algorithms for developing AI. Knowing which problems are easier to solve using Microsoft CNTK over other frameworks helps AI practitioners decide on the best software stack for a given application. In this course, you'll explore advanced techniques for working with Microsoft CNTK and identify which cases benefit most from MS CNTK. You'll examine how to load and use external data using CNTK and how to use its imperative and declarative APIs. You'll recognize how to carry out common AI development tasks using CNTK, such as working with epochs and batch sizes, model serialization, model visualization, feedforward neural networks, and machine learning model evaluation. Finally, you'll implement a series of practical AI projects using Python and MS CNTK.

Keras provides a quick way to implement, train, and evaluate robust neural networks in Python. Using Keras for AI development for prototyping AI is standard practice and AI practitioners need to know why and how to use Keras for particular AI implementations. In this course, you'll explore advanced techniques for working with the Keras framework. You'll recognize how Keras is different from other AI frameworks and identify cases in which it is advantageous to use Keras. You'll examine the functionality of the Keras Sequential model and Functional API and the role of multiple deep learning layers present in Keras. Finally, you will work with practical AI projects developed using Keras and troubleshoot common problems related to model training and evaluation.

Spark is a leading open-source cluster-computing framework that is used for distributed databases and machine learning. Although not primarily designed for AI, Spark allows you to take advantage of data parallelism and the large distributed systems used in AI development. AI practitioners should recognize when to use Spark for a particular application. In this course, you'll explore advanced techniques for working with Apache Spark and identify the key advantages of using Spark over other platforms. You'll define the meaning of resilient distributed databases (RDDs) and explore several workflows related to them. You'll move on to recognize how to work with a Spark DataFrame, identifying its features and use cases. Finally, you'll learn how to create a machine learning pipeline using Spark ML Pipelines.

The Amazon Machine Learning framework allows you to quickly deploy machine learning models using Amazon Web Services, automate model deployment and maintenance, and configure other Amazon tools to work in synchronicity. AI practitioners should consider the benefits and best practices of working with Amazon ML and other Amazon services in their AI development projects. In this course, you'll explore advanced techniques for working with the Amazon ML framework. You'll examine the significant differences between Amazon ML and other frameworks. You'll recognize the advantages of using the Amazon ML platform for certain projects and identify the Amazon ML workflow. Finally, you'll complete a project developing and training an AI model using the Amazon ML framework, and troubleshoot typical problems that come up during model training and evaluation.

The world of technology continues to transform at a rapid pace, with intelligent technology incorporated at every stage of the business process. Intelligent information systems (IIS) reduce the need for routine human labor and allow companies to focus instead on hiring creative professionals. In this course, you'll explore the present and future roles of intelligent informational systems in AI development, recognizing the current demand for IIS specialists. You'll list several possible IIS applications and learn about the roles AI and ML play in creating them. Next, you'll identify significant components of IIS and the purpose of these components. You'll examine how you would go about creating a self-driving vehicle using IIS components. Finally, you'll work with Python libraries to build high-level components of a Markov decision process.

Bidirectional Encoder Representations from Transformers (BERT), a natural language processing technique, takes the capabilities of language AI systems to great heights. Google's BERT reports state-of-the-art performance on several complex tasks in natural language understanding. In this course, you'll examine the fundamentals of traditional NLP and distinguish them from more advanced techniques, like BERT. You'll identify the terms "attention" and "transformer" and how they relate to NLP. You'll then examine a series of real-life applications of BERT, such as in SEO and masking. Next, you'll work with an NLP pipeline utilizing BERT in Python for various tasks, namely, text tokenization and encoding, model definition and training, and data augmentation and prediction. Finally, you'll recognize the benefits of using BERT and TensorFlow together.

Bidirectional Encoder Representations from Transformers (BERT) can be implemented in various ways, and it is up to AI practitioners to decide which one is the best for a particular product. It is also essential to recognize all of BERT's capabilities and its full potential in NLP. In this course, you'll outline the theoretical approaches to several BERT use cases before illustrating how to implement each of them. In full, you'll learn how to use BERT for search engine optimization, sentence prediction, sentence classification, token classification, and question answering, implementing a simple example for each use case discussed. Lastly, you'll examine some fundamental guidelines for using BERT for content optimization.

In this lab, you will perform AI Practitioner tasks such as performing gradient descent and stochastic descent, as well as baysean optimization. Then, test your skills by answering assessment questions after normalizing Tensor using Keras, training and evaluating Keras model, and extending Spark and the Markov Decision Process.

Final Exam: AI Practitioner will test your knowledge and application of the topics presented throughout the AI Practitioner track of the Skillsoft Aspire AI Apprentice to AI Architect Journey.