Erlend Øien
Evidi | Data & Analytics Consultant
Oslo, Norway
Actions
Erlend Øien is a Data engineer and Scientist, working as a Consultant at Evidi with an MSc in Computer Science from NTNU, Norway. His academic background includes a focus on databases, search, and recommendation systems. Erlend's skill set is a recipe of 3 cups data engineer, 1 cup software developer, 4 tablespoons data scientist, 2 teaspoons recommendation systems enthusiast, and a sprinkle of UX appreciation .
Erlend has dabbled with ML technologies for recommendation systems, explainable AI, as well as scaling workloads for distributed data processing. He has recently dedicated his efforts to data platform projects on Microsoft Fabric, with PySpark as his go-to tool. Moreover he is certified as an Azure Data Engineer, Data Scientist, and Fabric Analytics Engineer.
When Erlend is not refreshing the Microsoft Fabric blog for new features, he is passionate for everything regarding fermentation (and is edible or drinkable), particularly within the realm of baking!
Area of Expertise
Topics
Excel: Schrödinger's metadata database
How and where to define metadata for a metadata-driven framework often feels like choosing between lesser evils: verbose YAML files that punish bad indentation, sprawling JSON blobs, or fragile SQL tables only touched by brave souls. Some teams go so far as to build custom apps to manage metadata, which often feel like relics from 2005.
But what if the solution is something already in everyone’s toolbox? Enter Excel—an often-disparaged tool that’s both beloved and cursed, yet undeniably familiar and accessible regardless of its database-status. In this session, we’ll explore how Excel can serve as a structured, accessible entry point for defining the required metadata for an end-to-end ELT pipeline following the medallion architecture—without needing to build a full app or ask business users to learn Git.
We'll demo how OneLake File Explorer and open mirroring allow seamless syncing structured Excel inputs, including robust metadata validation, without the complexity of custom apps. While Excel is a great starting point, certain metadata—like dynamic pipeline variables and execution and auditing logs—are better handled by Fabric Databases. We’ll illustrate how this hybrid approach keeps things scalable and robust.
Of course, making Excel work for this isn’t just “save-as CSV and hope for the best.” We’ll demonstrate how to avoid turning Excel into yet another workaround. Moreover, we’ll tackle real-world challenges, such as how to ensure metadata entries follow the required schema, validate them effectively, and manage metadata versioning, to ensure that your pipeline remains robust and fully traceable.
By the end of this session, you’ll be enriched with an alternative—albeit slightly controversial—metadata-driven framework that balances flexibility with structure, showcasing that Excel doesn’t have to be the villain in your data story. By integrating Excel, we enable domain experts and business users to take ownership of their metadata, enhancing collaboration without sacrificing data quality or governance.
Enabling Machine Learning Workflows in Microsoft Fabric: Beyond Power BI with Semantic Link
Microsoft Fabric has emerged as a powerful platform in the Machine Learning (ML) and AI space, addressing common challenges organisations face in starting with ML and putting models in production. In this session, we’ll explore how Fabric simplifies end-to-end Data Science workflows, focusing on Semantic Link, a transformative feature for SMBs and enterprises alike. By accessing and synchronising Power BI semantic models to OneLake through Semantic Link, businesses can unlock advanced analytics without needing an extensive data warehouse.
We’ll demonstrate how to implement the theoretical Data Science Process in Fabric. Starting with the ideation and hypothesis phase, we’ll showcase data discovery of Semantic models using OneLake Catalog and explore options for synchronising Semantic Models, highlighting trade-offs in integration approaches and access methods.
For data exploration and modelling, we’ll use SynapseML and AI Functions to apply built-in GenAI for product segmentation, translation, and supplier reliability classification, enabled by Azure OpenAI. Next, we’ll prepare data models with AutoML tasks like financial forecasting, showcasing its accessibility for new teams and efficiency for scientists. Finally, we’ll showcase a scalable product recommendation system based on financial data, applicable across industries.
To provide actionable insights, models must be effectively deployed. We’ll demonstrate how to integrate the model lifecycle with your data pipeline, ensuring predictions remain relevant while monitoring trends and performance. Without a consumable layer, however, these models can’t deliver value. We’ll show how to serve predictions through Power BI semantic models and reports while seamlessly integrating them into external applications using Fabric’s shortcuts, SQL, or GraphQL as well.
By the end of this session, you’ll learn to implement end-to-end ML workflows in Fabric, from hypothesis to deployment, and integrate insights into Power BI and beyond. Join us to see how Fabric simplifies ML workflows and drives impactful results with Semantic Link!
Enabling Machine Learning Workflows in Microsoft Fabric: Beyond Power BI with Semantic Link
Microsoft Fabric has emerged as a powerful platform in the Machine Learning (ML) and AI space, addressing common challenges organisations face in starting with ML and productionising models. In this session, we’ll explore how Fabric simplifies end-to-end Data Science workflows, focusing on Semantic Link, a transformative feature for SMBs and enterprises alike. By accessing and synchronising Power BI semantic models to OneLake through Semantic Link, businesses can unlock advanced analytics without needing an extensive data warehouse
We’ll demonstrate how to implement the theoretical Data Science Process in Fabric. Starting with the ideation and hypothesis phase, we’ll showcase data discovery of Semantic models using OneLake Catalog and explore options for synchronising Semantic Models, highlighting trade-offs in integration approaches and access methods.
For data exploration and modelling, we’ll use SynapseML to apply built-in GenAI for product segmentation, translation, and supplier reliability classification, enabled by Azure OpenAI. Next, we’ll prepare data models for low-code AutoML tasks like financial forecasting, showcasing its accessibility for new teams and efficiency for scientists. Finally, we’ll showcase a scalable product recommendation system based on financial data, applicable across industries.
To provide actionable insights, models must be effectively deployed. We’ll demonstrate how to integrate the model lifecycle with your data pipeline, ensuring predictions remain relevant while monitoring trends and performance. Without a consumable layer, however, these models can’t deliver value. We’ll show how to serve predictions through Power BI semantic models and reports while seamlessly integrating them into external applications using Fabric’s shortcuts, SQL, or GraphQL as well.
By the end of this session, you’ll learn to implement end-to-end ML workflows in Fabric, from hypothesis to deployment, and integrate insights into Power BI and beyond. Join us to see how Fabric simplifies ML workflows and drives impactful results with Semantic Link!
Leveraging Semantic Link in Microsoft Fabric for Powerful Machine Learning Workflows
As data scientists with backgrounds in machine learning and MLOps across various platforms, we discovered Microsoft Fabric and were intrigued by its potential to meet and enhance our needs. In this presentation, we want to take you on a journey through Data Science and Machine Learning in Fabric, showcasing why it's such a powerful combination. We’ll highlight how quickly and effortlessly you can set up a machine learning pipeline using Microsoft Fabric’s features, with a special focus on the Semantic Link functionality.
This feature is especially valuable for small to medium-sized businesses (SMBs) that might not have the resources for a complex data warehouse. Many SMBs perform data transformations and joins within the Power BI semantic model because they lack a full medallion architecture. Semantic Link allows these models to be moved into Delta tables in Fabric for advanced AI processing, providing significant benefits without the need for a comprehensive data warehouse.
During our session, we will demonstrate how Semantic Link in Microsoft Fabric synchronizes a Power BI semantic model—integrating data from multiple on-premises and cloud sources—into Delta tables in Fabric.
We will then show how to use SynapseML to apply GPT-4 to text data within the semantic model, illustrating how Fabric’s integration with OpenAI services enables powerful text analysis and insights directly within your data framework.
Next, we’ll explore using the AutoML tool FLAML for regression and classification tasks. We’ll highlight how FLAML automates machine learning processes, making them accessible to teams without extensive ML expertise.
Finally, we’ll demonstrate how to return the predictions of the machine learning models back into the Power BI semantic model for real-time updates and enhanced analytics in your reports. This seamless loop—from data ingestion, through advanced processing, back to actionable insights in Power BI—showcases Microsoft Fabric's practical and transformative capabilities for SMBs and larger enterprises alike.
Join us to see how transitioning to Microsoft Fabric can simplify your MLOps practices and elevate your data science projects. Our session will provide valuable insights into using Semantic Link and other powerful features of Microsoft Fabric for efficient and impactful machine learning workflows.
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