

Håkan Silfvernagel
Manager Data & AI Enablement, Microsoft AI MVP
Oslo, Norway
Håkan holds a Master of Science degree in Electrical Engineering and a Master’s degree in Leadership and Organizational behavior. He has also taken courses on university level in psychology, interaction design and human-computer interaction. He has 20 years’ experience of software development in various positions such as developer, tester, architect, project manager, scrum master, practice manager and team lead.
Håkan is Chairman of the Norwegian .NET User Group Oslo (NNUG) and is active as an Ambassador for Oslo AI, the local chapter for the global City.AI community. In addition, he is the co-founder of AI42, an online school for learning about AI and Data Science and the Azure User Group Sweden, a meetup focused on Azure.
Håkan is a Microsoft Most Valuable Professional (MVP) in AI and a Microsoft Certified Trainer.
Currently Håkan is working as Manager Data & AI Enablement at Miles AS, a Norwegian consultancy company.
Area of Expertise
Topics
Affective Computing – bringing humans and machines closer through emotions
Research have shown that emotions play an integral role in decision making, cognition, perception, learning and more. If we want our computers to be intelligent and be able to interact with us we need to ensure that they are able to recognize, understand and express emotions. This is the basic assumption of the field of Affective Computing. In this talk I will give an overview of Affective computing and how it can be applied in order to make our interaction with machines more suitable to us as humans.
First I will give an introduction to the field starting with established findings from the field of psychology on how we best can measure emotions.
Then I will describe how the field of Affective Computing has transformed from its origin in the 90’s until now when it is an established research field. I will highlight some of the technology enablers that has made Affective Computing a hot topic nowadays and give some examples of API and services that we as developers can use as of today.
In the second part of my talk I will give some examples on application scenarios across various fields (retail, medical, education and social). After that I will be show casing what is in the front line now. I will conclude my presentation with some recommendations on how this affects us as developers going forward.
Machine learning in the browser using TensorFlow.js
In order to start out with machine learning you typically would need to learn Python, Tensorflow, Jupyter Notebook etc. But what if you could run your machine learning straight in the browser. This can be done through Tensorflow.js. In this session you will get an introduction so that you can use it in your own projects.
This session will give you an introduction to what Machine learning is and what types of problem you can solve. TensorFlow as a library will be introduced and then TensorFlow.js will be presented with a focus on how you can use a machine learning model in your JavaScript application.
Next, we will build an image classification web app that uses a predefined TensorFlow model.
Finally, some examples on how TensorFlow.js is used in commercial applications will be given.
Machine Learning on the edge using TensorFlow Lite
What if you could perform machine learning on the edge, i.e on your mobile device? This would mean that you no longer would need the roundtrip to the server, no data will leave the device and you don't even need an internet connection . In this session you will get an introduction to TensorFlow Lite so that you can use it in your own projects.
In this presentation I will demonstrate how you can take a pre-trained model for image classification and convert it to TensorFlow Lite format. Then I will go through how to deploy to model to the device and finally talk a little bit about optimization options in order to reduce the model size.
The History of AI - what can we learn from the past?
Nowadays AI is all the hype, but what many might not know is that AI is an established discipline originating from a paper from Alan Turing in the 1950s. In this talk I will present the historical milestones of AI from the originating paper up until present days. In addition we will look into the crystal ball in order to see what the future might have in store.
We will start out our journey by looking at what happened in a workshop in Dartmouth in the 1950’s which started it all. Then we’ll be reviewing a number of areas where AI initially was put to use between 1950-1970. We’ll cover the AI winter in the 1980’s and its’ reasons.
In the second part of the talk we’ll cover applications and milestones from the 1990’s and onwards. Finally we’ll look into the crystal ball and try to see where AI might takes us in the future.
Why you should consider Web Assembly in your next frontend project
During the last decades a growing trend has been to put more and more functionality into the client by using the latest and greatest JavaScript framework. But what if we could be using native code in the browser in order to run computations faster and potentially reuse code from the backend in the frontend.
Enter Web Assembly. Web assembly is a new web standard which enables you to run native code as part of your current JavaScript framework. This talk will give you a thorough understanding of what web assembly is and how you can use it in your project.
We will cover a practical example writing our web assembly using Rust. We will go through everything from writing your web assembly code to publish it as a npm package and finally use it in an existing web application.
What's the big deal with TinyML
Presently there are billions of IoT devices in use ranging from alarm systems, smart fridges and clocks to thermostats. Having many devices communication over the networks simultaneously may lead to saturation of the bandwidth. This has generated interest in a new paradigm for computing: the data-centric paradigm. Instead of bringing data to the computing power, we bring the computing power to the data.
This means that there is huge potential for performing machine learning on small embedded systems like microcontrollers. This brings some challenges as most microcontrollers have limitations in terms of both memory and computing capacity. Enter TinyML.
In this session you will learn about what TinyML is, what kind of problems it is suitable for and a demonstration of a practical example of developing, testing, training and deploying a TinyML model onto a physical hardware. I will be demonstrating the workflow for working with TinyML, demonstrating how we can convert a trained TensorFlow model to TensorFlow Lite and how we can deploy the TinyML onto hardware.