Nordics AI Summit schedule

While the deployment of artificial intelligence systems is seen as a pioneering and groundbreaking arena, their commercialization and innovation require an ethical filter. 
No dataset is bias-free. It is entangled in the process of the many organizational silos between technical, legal, ethical, and business departments, and the lack of effective control management across these departments.

Biases are inherent in us human beings. We cannot avoid them. However, we can try to do more to govern AI. Ungoverned AI such as data bias, misuse/over-use of data and AI and immature data can lead to costly reputational harm and lack of trust.  All organizations must be accountable for how their use of data and AI is affecting people and society. The European Union is expected to introduce a regulatory framework for Artificial Intelligence in 2024, which will require AI reporting and new governance structures. This means that organizations must consider ethical AI across all functions within the organization. Compliance and regulation alone cannot be relied on to do the right things. It is a starting point, and having a sound AI framework moving forward is essential. 

Jonathan Rebane is an AI and Machine Learning Expert at anch.AI where he brings technical expertise for developing sustainability oriented AI methods, solutions, and data insights. He is passionate about sustainable AI and believes that various stakeholders must work together to ensure responsible and trustworthy data driven innovation.

Jonathan has experience in multiple areas of academic research and the private sector working in Data Analytics. Jonathan completed his PhD in Data Science at Stockholm University. His research involves developing AI frameworks to predict future diagnoses for patients using Electronic Health Record Data. In addition, Jonathan has published works on ”Explainable and Interpretable AI”,  ”AI for Cryptocurrency Price Prediction”, ”Mathematical Modelling in Cellular Neuroscience” and ”Spatial Cognition.” A complete list of Jonathan’s publications can be found here.

Jonathan holds a Master’s degree in Health Informatics from Karolinska Institutet & Stockholm University and a Master’s degree in Neural Information Processing from the International Max-Planck Research School for Cognitive and Systems Neuroscience in Germany.

Husqvarna Group (founded in 1689 in Sweden) is a leading manufacturer of, for example, forest and garden equipment, watering products, as well as machinery for the construction industry. This presentation will give a brief overview of some use cases (without sharing details of data/results) where deep learning could be applied in the R&D of various products and services. For example, deep learning-based instance image segmentation (eg. MaskRCNN) was applied to understand a chainsaws cut process through wood. In other use cases, time series data was used for predicting supply-chain demand. Another example will show how machine learning could be applied to the understanding of the injection molding process.

Nasir Uddin' PhD is a Senior data scientist and AI/machine learning engineer with a multidisciplinary educational background who can combine research and business together.
He is good at programming in Python and R as well as working with various open sources AI frameworks and cloud computing platforms including Azure, Google cloud, AWS, IBM etc. He is also familiar with SQL, Javascript, Java, Swift as well as AI software/application development, data engineering and big data architecture within industry 4.0 concept. Nasir worked at various agencies/enterprises/industries dealing with finance/banking, energy, infrastructure, real estate, environment, space/satellite, car manufacturing, Food-Pharma and process. He holds PhD in Genetics from Aarhus University in Denmark, MSc from Uppsala University in Sweden and vocational certificates in Software/Mobile Application Development from Nackademin in Sweden.

 

The most popular data science development tools have largely been developed by academics as scratch pads for interactive data exploration.  Jupyter notebooks, for instance, were developed 20 years ago at Berkeley (they were called iPython notebooks at the time).  Because of their flexibility and interactivity, these tools have become widespread amongst coding data scientists.  More recently, GUI-based tools have begun to be popular.  They reduce the technical load on the user, but typically lack much needed flexibility and interoperability.  Both avenues of innovation are wildly inadequate for modern data science development.  GUI-based tools are typically too expensive, too restrictive, and too closed.  The development of automated machine learning tools only made this problem worse, with dozens of software startups urging business analysts to start building machine learning solutions, often with questionable results and even more questionable customer retention metrics.  On the other hand, notebook-based solutions are typically too error-prone, too loose, and too isolated to be sufficient.  The result is intractable challenges around collaboration, communication, and deployment.  The most recent entrants into the notebook space have only marginally improved the experience without fixing the underlying flaws.  This talk discusses the fundamental flaws with the way these tools have been developed and how they currently function.  Advancement in this space will require reworking the architecture and functionality of these tools at some of the most basic levels.  These fixes include things like multiprocessing capabilities; real-time collaboration tools; safe, consistent code execution; easy API deployment; and portable communication tools.  Future innovation in the data science development experience will have to tackle these problems and more in order to be successful.

Greg Michaelson is Co-Founder and Chief Product Officer at Zerve, a young, stealthy startup that’s rethinking the data science development experience. Previously, Greg was an early joiner at DataRobot where he played many roles, including Chief Customer Officer. Prior to that, he worked as a data scientist in the financial sector after earning a PhD in Applied Statistics from the University of Alabama.  In his spare time, Greg manufactures a line of flavored breakfast cereal toppings called Cerup. He lives in Spring Creek, Nevada with his wife, four children, and two Clumber Spaniels.

Training a great machine learning model isn’t enough – it isn’t useful until you deploy it. In this talk, the lead developer of SEB’s MLOps platform candidly shares how it was built, what worked well and less well, and how the shift to a cloud-based solution is changing things.

Manne's experience includes backends for high traffic websites in the gaming and media industries, telecom systems and IPTV. He's talked at several software development conferences and has often worked as a mentor and teacher. He has deep knowledge in Java, Test-driven development, Domain-driven design and server architecture and is fascinated by the new possibilities offered by Big data and machine learning algorithms.

When we in the ML-space talk about scaling machine learning, the focus usually falls on the technical aspects such as memory requirements, latency and throughput. When a company grows, and its machine learning applications grow with it, another aspect of scaling arises - an organisational one. A growing number of use-cases for ML combined with a growing team requires an ML-Ops based infrastructure, enabling both technical scaling and efficient ways of working. At Babyshop Group, an e-commerce scaleup, we have 10+ productionised ML-based applications within a relatively small team, made possible by leveraging ML-Ops best practices. During this talk, I will share our infrastructure, key learnings, and how we can sustain a high rate of innovation at scale.

 

Marcus Svensson is the Head of Data Science at the Babyshop Group, where he is responsible for the data science team (3 people), leveraging data across all stages of analytics, from descriptive to predictive to prescriptive, in order to create value across the whole organisation. Use-cases for in-house ML-based solutions involve; customer insights & segmentation, CLV-predictions (for Google Ads), attribution modelling, recommender systems (on-site / email), smart sorting (listing pages & search), demand predictions as well as intelligent pricing. In order to support such a wide range of use-cases, the underlying tech-stack and infrastructure used are a combination of Kubernetes & Serverless hosted on GCP, with emphasis on containerisation. More specifically a combination of Kubeflow Pipelines (ML-pipelines running on Kubernetes), Cloud Run (serverless REST API’s), Cloud Functions (serverless jobs) as well as Vertex AI (distributed hyper-parameter tuning). 

Today’s larger AI models demand both extreme compute capacity, time-consuming and costly pre-work. But this is not enough; in addition, you must make substantial trade-offs in splitting the models and dealing with the sparsity dilemma to run on today’s standard CSP structure – even with a GPU based cloud provider. Finally, the staggering energy consumption is far from sustainable, neither from a climate nor from a cost perspective.
To meet these challenges, Green AI Cloud will present a cloud solution based on the Cerebras Systems.

Jacob Boström has a master’s degree in computer science from the Stockholm Institute of Technology (KTH). He has more than 15 years of experience of building leading edge telecom/AI platforms for Ericsson globally. This includes extensive programming experience and track record, primarily within telecom, where he has delivered many complex projects. One of the larger projects was Ericsson’s login-system IAM which today is used by some 100 000 employees around the world. Also worth mentioning is the groundbreaking work for Ericsson research department developing an AI simulation platform with cloud architecture. Jacob is the founder of Green AI Cloud. His earlier venture – Fredenheim a high profile data consultancy company- was nominated” Super company” both 2017 and 2018, by one of Sweden’s biggest business magazines.

André specialized in designing and deploying ML solutions on resource-limited edge devices, currently working at Voi to make e-scooters safer. Prior to this he was a Machine Learning Advisor working with data collection, predictive maintenance and computer vision at various companies including Metso, Volvo GTT and CEVT.

There are many approaches to extract feature vectors/embeddings from the samples (especially more about vision samples here) in both conventional and learning based, am going to talk the different neural network designs and their efficiency improvements (in terms of both accuracy and computational time) starting from classification networks to 2D and 3D detection architectures. How feature extractor is important in both 2D and 3D detections and different designs of head networks with different branches to produce different outputs.

Veeranjaneyulu Toka is a  Computer Vision Research Engineer at Hyke. With 15 Years of experience in the Multimedia domain and vast experience especially in Camera and Image/Video solutions.
His work focuses on Research & Development in Computer Vision and Machine Learning, improving accuracy and performance, robustness and productization in memory and performance constraint devices, extensive problem-solving and coding skills, and he often participates in HackerRank contests.

 

Prediction of a varying number of ordered clusters from sets of any cardinality is a challenging task for neural networks, combining elements of set representation, clustering and permutation learning. This task arises in many diverse areas, ranging from medical triage, through multi-channel signal analysis for petroleum exploration to product catalog structure prediction. This presentation focuses on the latter, referred to further as the eponymous Catalog Problem, which exemplifies a number of challenges inherent to adaptive ordered clustering. These include learning variable cluster constraints, exhibiting relational reasoning and managing combinatorial complexity.

Mateusz is a Senior Machine Learning Engineer at Tjek A/S, a Danish company helping people with their shopping via their eTilbusavis app. His current focus is on deep learning solutions to structure prediction, set representation and neural clustering, which is also the subject of his Industrial PhD thesis. Prior to his current position he worked at Intel, WNS Global and Procter and Gamble. 

 

By taking neural networks back to the school bench and teaching them some elements of geometry and topology we can build algorithms that can reason about the shape of data. Surprisingly these methods can be useful not only for computer vision – to model input data such as images or point clouds through global, robust properties – but in a wide range of applications, such as evaluating and improving the learning of embeddings, or the distribution of samples generated by GANs. This is the promise of the emerging field of Topological Data Analysis which we will introduce and review recent works at its intersection with machine learning.

 

After a career as entrepreneur and data scientist, Jens turned to research and is currently pursuing a PhD at KTH within the WASP program. He believes that an important property that sets humans apart from other animals is that we have a sense of geometry and topology. Teaching computers a sense of geometrical recognition and reasoning is thus a promising direction if we want to develop more powerful AIs. To do so we must expand the mathematical toolkit underlying machine learning to include math from less well-known fields such as computational geometry and topology. Jens will discuss some of the results and challenges in doing so. 

With an ever-growing market of ML/AI/DS tools, it can be a jungle to find the right ones. Most use cases are seeing end-to-end products that can be deployed on cloud resources. Nevertheless, having full control of the deployment hardware is not always easy as with edge computing or your optimization needs might stretch beyond the current sea of python libraries. Can we interface compiled languages with python, as TensorFlow or NumPy do it? Should this be done by the data scientist, the data engineer, the ML engineer, or a subset of these in collaboration?

Morten is currently the Head of Advanced Analytics at DNB Wealth Management, responsible for the machine learning models used for customer insight and prediction. He is also responsible for the operational data governance activities in our business area.

He studied at the Norwegian University of Science and Technology (NTNU) and received an M.Sc in applied mathematics in 2001 and started my career in Norsk Hydro, working with energy trading and estimating the value of oil fields. In 2007 he started as a quantitative portfolio manager for global equities in DNB. Since then, he’s had various positions within equity trading and portfolio management in DNB, always with a focus on programming, statistics and machine learning.

In 2018 he started the Advanced Analytics department in DNB Wealth Management. It started out as a small team initially and has grown in size as our responsibilities expanded. They are now a team of 10 data scientists, data analysts and data stewards

Data Scientist Leader, multilingual, seeking to drive impact and build a world class data science team at ICA. A skilled communicator capable of explaining complex issues with clarity to diverse audiences. Interested in growing data talent while creating an inclusive environment and developing scalable processes to make data accessible across organizations, especially to non-technical stakeholders.

Anna’s ambition is to be an inspiring leader who empowers and builds trust within her team. Highly adaptable and result-driven, she has demonstrated valuable leadership skills in technically complex environments, through periods of rapid change and in high-pace environments. Strong analytical skills and fast processing of complex information enables me to stay calm and methodical in all situations.

Janne is a leader specialized in developing data driven organizations and strategies, with hands-on expertise in advanced analytics. A highly versatile professional, with experience from leading client engagements as well as corporate level transformation initiatives in complex matrix organizations. He is a recognised thought leader and frequent speaker around digitalization and AI within the industry.

Maria is Head of Artificial Intelligence Program at SSAB, where her focus is on leveraging data sciences and machine learning solutions to bring value to the business.

She has long experience in the steel industry, working on bringing together business needs and IT possibilities to create solutions for customers or improve own company processes.

She has an Engineering degree and holds a PhD within the field of Energy Systems. She has also been honing her skills in Statistics & Data Sciences.

How can we use data to support sustainable digitalisation? The effects of digitalisation on sustainability? When we work on AI systems, during training and deploying ML/DL models, the carbon footprint is huge - what are some sustainable practices to reduce the carbon footprint when training/ deploying AI models? What are some common pitfalls of sustainable digitalization, and how can we successfully avoid the pitfalls? I will also share use cases from my experience working with AI for sustainable development.

 

Smriti is passionate about finding patterns and making predictions from empirical data that not only revolutionize the day-to-day work process but also help humanity overcome its most critical challenges such as climate change, healthcare challenges and the problem of lack of diversity and inclusion within technology.
Previously, she worked as the head of AI at Earthbanc where she was advocating for innovation in the fight against Climate Change using her knowledge of tech and AI. She worked on creating a dynamic, reliable and accessible Carbon Monitoring, Reporting and Verification (MRV) model which would help make the MRV service accessible to landowners with a small area of land (patent-pending).

Fundamentals of Responsible AI Models 

We all know that technology amplifies the existing entanglement of social constructions, relations, and practices. If the perspectives of marginalised users are not considered in AI development, implicit biases and damaging practices will persist. But how do we implement responsible models in practice? In this session, let's talk about the fundamentals of responsible AI at all stages of development process, from data collection to model performance monitoring.

Irina has 15+ years of experience as a data scientist and policy expert as well as a professional coach, director, and mentor. She has a PhD with a track of publications in political economy, investment analysis, and international relations. 

At Ahold Delhaize, we have multiple brands in different countries and many of them want to have similar ML models running on their websites. We would like to show, how we create a system that allows us to deploy the same model for different brands, without reinventing the wheel.

The components we developed for reusability:
- Configurable central python package for each model
- Cookiecutter template to create repositories with necessary components
- Reusable pipelines to deploy models
- Identical infrastructure& service user setups

 Maria is a Senior Machine Learning Engineer at Ahold Delhaize. Maria is bridging the gap between data scientists infra and IT teams at different brands and focuses on standardization of machine learning operations across all the brands within Ahold Delhaize.
During more than eight years in Data&Analytics, Maria tried herself in different roles, from data scientist to a machine learning engineer, was part of teams in various domains, and have built broad knowledge. Maria believes that a model only starts living when it is in production. For this reason, last six years, her focus was on the automation and standardization of processes related to machine learning.

At Ahold Delhaize, we have multiple brands in different countries and many of them want to have similar ML models running on their websites. We would like to show, how we create a system that allows us to deploy the same model for different brands, without reinventing the wheel.

The components we developed for reusability:
- Configurable central python package for each model
- Cookiecutter template to create repositories with necessary components
- Reusable pipelines to deploy models
- Identical infrastructure& service user setups

Basak Eskili is a Machine Learning Engineer at Ahold Delhaize. She is working on creating new tools and infrastructure that enable data scientists to quickly operationalise algorithms. She is bridging the space between data scientists and platform engineers while improving the way of working in accordance with MLOps principles.
In her previous role, she was responsible for bringing models to production. She focused on NLP projects and building data processing pipelines. Basak also implemented new solutions by using cloud services for existing applications and databases to improve time and efficiency.

Leonard is currently driving ML Engineering and Operations at Volvo Cars. He’s responsible for defining the overall mission and strategy for ML Engineering and Operations, leading the build of maintainable ML systems. He has hands-on ML Engineering & Data Science expertise. He’s passionate about making the world a better place with ML and doing great things with data.

Founded almost 80 years ago, IKEA has become the world largest home-furnishing retailer in the world, always led by its values. The values can be seen in the actions that are taken, the range offered and how IKEA strive to create a better everyday life for the many people. Being a values-led brand, we are guided by our values in a digital context and when making decisions around data and technology. One of the keys for success within Inter IKEA Group is the mindset and approach around values-based decision-making and focusing on creating a positive impact on people, the planet and society. Algorithmic and AI Ethics therefore becomes a natural part of how we live IKEA culture and values in the digital age.

Giovanni Leoni is the Global Head of Algorithmic and AI Ethics at Inter IKEA Group, responsible to develop, implement and lead operationalization of responsible analytics across Inter IKEA Group of companies. This concerns all algorithmic processes, including administrative processes, operational processes, analytics and AI. He is an active contributor to the global AI ethics community, engaging in legislative industry consultations and various forums and bodies, among them being on the Board of Directors for Ethical Artificial Intelligence Governance Group. For more than 20 years he has been active across multiple sectors, private and public, driving development of user experience, sustainable change management, clarity in impacting factors for better business and wrapping everything in numbers and decision-models. He has worked within diverse areas as public sector processes, sales, procurement of services, digital and physical products, supply chain, automatization and business analytics. In his current assignment it all comes together, being a partner to create better business with the use of data through values-led decision-making.

 

Autonomous vehicles are one of the most hot topics these days. The complexity required to solve autonomous driving makes it inevitable that in some parts of the problem, AI will be used. It can be used in many sections like perception, planning, prediction, may be even end to end.
At research and development at SCANIA we have been investigating the application of AI for heavy-vehicle autonomous transport, and in this presentation, we will have a 10000 foot view of some of works that have been done in the society and also a snapshot of some of the things we are doing within the area of perception using AI.

Shahab (M.) has finished his Ph.D. in 2017 and joined SCANIA after that. Since then, he has worked different fields involving machine learning and AI at SCANIA.

Maddy is Experienced in leading multifunctional data science and machine learning teams and is experienced in advanced data, analytical and machine learning techniques, and supporting product development with data-driven decision making. She has a PhD in image processing, computer vision and machine learning with computing science background, experienced in both industry and academic research. Her main work is in mobile gaming, social network, image applications in both medical and industrial data.

Daniel is a Machine Learning Researcher at the RISE Institute of Sweden where he focuses on applying machine learning and probabilistic models to develop intelligent autonomous systems for solving combinatorial optimization problems for networked systems.

 Irina has 15+ years of experience as a data scientist and policy expert as well as a professional coach, director, and mentor. She has a PhD with a track of publications in political economy, investment analysis, and international relations. 

Covid-19 accelerated the digitalization of retail. To meet the increased demands for instant and high-quality service, contact centers around the world invested in AI to empower their human workforce. In this session, Anna Hjalmarsson, lead AI engineer at Electrolux, will share experiences from business transformation and AI democratization when building an AI powered omni channel contact center. 

Anna is a specialist in conversational AI, currently working with Electrolux building a new platform for Customer Services