Alan Smith
AI Developer, Active Solution
Stockholm, Sweden
Actions
Alan Smith is a Cloud & AI Trainer, Mentor & Coach at Active Solution in Stockholm, Sweden. He has a strong hands-on philosophy and focusses on embracing the power and flexibility of cloud computing to deliver engaging and exciting demos and training courses.
Alan has held the MVP title since 2005, and is currently an AI MVP. He is in the organization team for the CloudBurst and AI Burst Conferences.
Links
Awards
Area of Expertise
Topics
Real World Vision Solutions using Azure Computer Vision, Azure Custom Vision and Azure ML
Creating AI based vision solutions can in some cases be a challenging task, requiring days of training time using complex neural networks with large image datasets on expensive GPUs. In other cases, solutions can be created by making a simple API to call Azure Cognitive Services such as Computer Vision and Custom Vision. Understanding the strengths and capabilities of the Azure AI services along with the challenges and best practices of developing AI based vision solution will enable you to choose the appropriate strategy for solving vision-based AI challenges.
In this demo intensive session Alan will use real-world scenarios to demonstrate the strengths and capabilities of different vison solutions, ranging from image classification for an auction site to object tracking from drone video footage. For each scenario he will discuss the challenges involved and highlight the appropriate technologies that can be used to create a working solution. Alan will also share his experience and opinions on when to use Azure Cognitive Services vs. a self-trained model, provide guidance on getting to grips with ML frameworks such as PyTorch, and how Azure ML can be used as a platform for AI development.
Inside GPT – Large Language Models Demystified
Natural language processing using generative pre-trained transformers (GPT) algorithms is a rapidly evolving field that offers many opportunities and challenges for application developers. But what is a generative pre-trained transformer, and how does it work? How can you leverage the latest advances in GPT algorithms to create engaging and useful applications? Can my business benefit from creating a GPT powered chat bot?
In this demo intensive session Alan will take a deep dive into the architecture of GPT algorithms and the inner workings of ChatGPT. The journey will begin by looking at the fundamental concepts of natural language processing, such as word embedding, vectorization and tokenization. He will then demonstrate how you can apply these techniques to train a GPT2 model that can generate song lyrics, showing the internals of how word sequences are predicted.
Alan will then shift the focus to larger language models, such as ChatGPT and GPT4, demonstrating their power, capabilities, and limitations. The use of hyperparameters such as temperature and frequency penalty will be explained and their effect on the generated output demonstrated. He will then cover the concepts of prompt engineering and demonstrate how Retrieval Augmented Generation (RAG) patterns can be leveraged to create a ChatGPT experience based on your own textual data.
Join me for this session if you want to learn how to harness the power of GPT algorithms in your own solutions.
Building an Image Similarity Search using Spotify Annoy, PyTorch and Azure Machine Learning
Spotify Annoy (approximate nearest neighbor Oh Yeah!) is an open-source algorithm used by Spotify for identifying similar sounding songs for recommendations to users. Spotify Annoy can also be used to create a search index for similar images, which has many real-world implementations including recommending products in on-line stores. Creating a similar image search index can be accomplished in a few lines of Python code, but how can this process be automated, and the index published as an API that can be consumed by other applications?
In this demo intensive session Alan will run through the process of creating an image similarity search API hosted in Azure Machine learning. Starting with the creation of an image dataset he will create an Azure ML experiment to use a pre-trained PyTorch model to create an approximate nearest neighbor index using Spotify Annoy. He will then create an endpoint in Azure ML that return images that are similar to a target image. Throughout the process he will explain the theory of using PyTorch and Spotify Annoy and how the features of Azure ML Studio can be leveraged for the rapid cloud-based development of machine learning solutions.
Chat with your Structured Data with ChatGPT, SQL and More
Developers can leverage ChatGPT or GitHub Copilot to generate sophisticated SQL queries to retrieve relevant results from a database from a text prompt. Why not create a chatbot with the ability to generate and also execute SQL queries against a database? Providing the option to access structured data opens up the possibility for a chatbot to answer complex questions relating to business data using a natural language interface.
In this demo intensive session Alan will explain concepts of integrating language models with structured data sources such as SQL Server. He will demonstrate how SQL chains and agents in LangChian can be leveraged to connect to a database, analyze the table structure, and then compose and execute the queries used to generate the results. The internals of SQL chains and agents will be covered along with tips about prompt optimization. Alan will also cover the safety and security aspects of allowing an AI to access external data.
GPT-4 vs Starcraft II – Strategic Decision Making using Large Language Models
Starcraft II requires complex strategy decisions to be made in real-time. Planning resources, building the optimal structures and units, when to upgrade and when to defend or attack will all be critical to the success in a game. Large language models, such as GPT-4, have shown impressive performance on various natural language tasks, such as text summarization, question answering, and text generation. But how well can they make strategic decisions in a dynamic and competitive environment?
In this demo intensive session Alan will explore the capabilities of large-language models for strategic decision making. He will explain the strategy decisions that need to be made in a Starcraft game, and what makes it an ideal scenario for exploring and evaluating the capabilities of GPT-4. Alan will then focus on the techniques for leveraging GPT models for strategic decision making, including prompt engineering and state description as well as parsing and understanding the response messages. He will also discuss different scenarios where large language models can be leveraged in strategic decision making.
Join me for this session if you want to learn more about using GPT models in strategic decision-making processes, or just sit back and watch GPT-4 destroy the Zerg.
Build Natural Language Solutions with Azure Open AI Service
Azure Open AI Service is a new cloud-based platform that enables developers to create natural language solutions using state-of-the-art models such as GPT-4 and Codex. The Service provides a simple and intuitive interface to access these models, as well as tools to customize, optimize, and monitor their performance. Whether you want to build a chat bot, a code generator, a text summarizer, or any other natural language application, Azure Open AI Service can help you achieve your goals.
In this demo intensive session Alan will show you how to use Azure Open AI Service to build natural language solutions from scratch. He will start by explaining the basic concepts of natural language processing, such as tokens, embeddings, and transformers. He will then demonstrate how to use the Azure Open AI Service portal to create and deploy natural language models using pre-trained or custom data. He will also show you how to use the Azure Open AI Service SDK to interact with the models programmatically and integrate them with other Azure services.
Join me for this session if you want to learn how to leverage the power of Azure Open AI Service in your own solutions.
Open AI & GPT Workshop
Introduction
Large language models (LLMs) have rapidly transitioned artificial intelligence from almost obscurity into the mainstream. The constantly evolving models and the impact that they have on society have recently become front-page news. Many developers, data scientists and companies are rushing to adopt and leverage these developments in their business process, products and solutions.
• What do I need to consider when developing an LLM based solution?
• What is a generative pre-trained transformer, and how does it work?
• How can these models be leveraged by application developers to deliver real business value?
In this workshop you will explore how to leverage the OpenAI and 3rd party models to develop AI enhanced solutions. You will learn options for hosting and consuming different models and develop code and prompts to explore their capabilities. The integration with company knowledgebases to provide a ChatGPT experience tailored to your business requirements will be covered along with the concepts of using prompt engineering to influence the outputs generated by the model.
Hands-on labs using the latest LLM frameworks will be available in C# and Python, ranging from walk-through exercises to advanced challenges and group activities.
The field of LLMs is constantly evolving, the contents of the workshop are being continuously updated to cover the latest developments.
This is what you will learn
• Large Language Models: Learn about the history, current state and possible future of the rapidly changing field of LLMs, focusing on hosted and open-source models.
• OpenAI, Azure OpenAI & Azure Machine Learning: Understand the different capabilities of cloud-hosted LLM model offerings, including options for model selection and fine-tuning.
• Prompt Engineering: Explore and experiment with the different techniques of prompt engineering, including techniques for output formatting, jail-breaking and securing your applications.
• Developing LLM Solutions: Leverage frameworks such as LangChain, Semantic Kernel and Prompt Flow to develop solutions that integrate with LLMs.
• Inside GPT: Gain an understanding of the internals of GPT models and how tokenization, embedding and output sampling work together with the model’s attention mechanism.
• Retrieval Augmented Generation (RAG): Integrate GPT solutions with external services, such as vector-based, text and hybrid search engines, to develop “chat with your data” solutions.
• Testing & Evaluating Responses: Understand the challenges of testing and evaluating the responses generated by LLMs and develop a testing strategy that provides quantitative metrics of the output quality.
• Working with Agents & Plugins: Explore the power of using pre-built and custom agents and plugins to create tools that LLM solutions can leverage to perform tasks and integrate with other systems.
Who should attend?
This workshop is ideal for developers and data scientists looking to deepen their understanding of generative AI and LLMs and integrate them into their projects.
Programming experience in C# or Python will be required for most of the hands-on labs.
At the end of this workshop, you will be able to:
• Understand the evolution and internals of LLMs
• Utilize cloud-hosted and 3rd party models
• Master prompt engineering
• Implement retrieval augmented generation (RAG)
• Test and evaluate LLM responses
• Leverage and develop agents & plugins
What should you bring?
• Laptop with a development environment suitable for Python or C# development.
• Access to OpenAI or Azure OpenAI services.
AI Assisted Development with Chat GPT and GitHub Copilot
Large Language Models have evolved quickly and are revolutionizing the way people work and engage with AI. Models such as Chat GPT and GitHub Copilot can generate code in almost any programming language and identify bugs and issues in code. Leveraging the power of these models to improve developer productivity is currently a strong focus for many companies and developers.
AI assisted code generation also raises important issues and questions.
• Can AI generated code be trusted to function correctly?
• Will AI replace developers in the workforce?
• How can AI be leveraged in the best way to improve productivity?
This workshop will provide hands-on experience of leveraging Chat GPT and GitHub Copilot to assist in the software development process. AI will be leveraged to assist with the following tasks:
• Brainstorming ideas and suggesting features.
• Generating code for a prototype application.
• Creating test code.
• Identifying issues and fixing bugs.
• Generating images and UI graphics.
There are many different scenarios that could be used for the hands-on implementation, some examples are provided below.
Integrating with Azure Services
The Azure platform provides an extensive range of cloud-based platform as a service (PaaS) services enabling developers to build scalable and reliable applications. Starting with a simple web application, you will leverage Chat GPT and GitHub Copilot to extend the application to integrate with other Azure services to enhance the functionality.
Game development in Python
The PyGame library provides a quick and productive entry to developing basic games in Python. The extensive range of sample code in the public domain means that Chat GPT and GitHub Copilot are able to generate usable code for a starting project. This can then be extended using generated code to enhance the game, trouble shoot issues and develop and test features. Generative AI can also be used to create graphics and artwork for the game.
File -> New -> Project…
Chat GPT and GitHub Copilot are excellent tools for learning a new programming language, improving your programming skills and learning new techniques or SDKs. Feel free to choose your own scenario using Chat GPT and GitHub Copilot to accelerate your learning experience. When selecting a scenario, it is best to focus on exploring a new area, technology or language.
Alan Smith
AI Developer, Active Solution
Stockholm, Sweden
Links
Actions
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top