Serverless ML in Action: Real-World Projects from KTH

Kicking off the year with our first meetup, we are continuing the tradition of showcasing real-world serverless ML projects built by KTH students!

From forecasting cross-border energy flows to predicting transit delays with live weather data, these projects highlight how fully serverless ML systems can tackle real-world challenges.

Agenda:

17:30 - 18:00: Doors open
18:00 - 18:20: Welcome - Intro
18:20 - 18:35: E-FlowCast – Forecasting Cross-Border Energy with Serverless ML
18:35 - 18:50: Real-time weather-based transit delay forecasting
18:50 - 19:20: Pizza & Beers
19:20 - 19:35: The Ultimate 1Million Movies Dataset
19:35 - 19:55: JobsAI
19:55 - 20:30: Networking

–
Introduction
Fabian Schmidt, PhD Student, KTH

Fabian is working on the ALEC2 project under the supervision of Professor Vladimir Vlassov and Associate Professor Amir Hossein Payberah at KTH Royal Institute of Technology.

Presentations:

E-FlowCast – Forecasting Cross-Border Energy with Serverless ML
Rosamelia Carioni - KTH Student
Eric Banzuzi - KTH Student

How do you build a serverless ML system that predicts real-time electricity flows? The energy grid is one of the world's most complex and volatile systems. Prices fluctuate, demand spikes and renewable sources make forecasting difficult. To stay ahead, traders, operators, and policymakers need reliable predictions. E-FlowCast is a serverless ML system designed to address one area of this challenge, providing real-time electricity flow predictions for the next 24 hours.

Speakers Bio:
Rosamelia is a final-year master’s student in Machine Learning at KTH. She has a BSc in Data Science and AI from Maastricht University. Through research and internships, she has worked on predicting donor churn, building an NLP-to-SQL chatbot to improve data accessibility, developing an energy-weather visualization tool for real-time trading insights, and enhancing gunshot detection for wildlife conservation.

Eric is a final-year master's student in Machine Learning at KTH, with interests and skills in ML, data engineering, and software development. Prior to KTH, he completed a BSc in Data Science and AI at Maastricht University in 2023.

Real-time weather-based transit delay forecasting
Paul Hübner - KTH Student
Jonas Müller - KTH Student

When snow falls, there are often drastic consequences for the public. Public transportation of all types experiences sudden large delays, leading to people being stuck waiting at stops and missing transfers. By taking local weather forecasts into account, such travel delays could be predicted in real-time. Based on historical and live transit delay data from Gävleborg County as well as Open-Meteo forecasts, we present a real-time serverless delay forecasting dashboard powered by machine learning.

Speakers Bio:
Paul studied Computer Science & Engineering at TU Delft, specializing in the Machine Learning and Data Science track. He is pursuing further studies in Software Engineering of Distributed Systems at KTH.

Jonas is pursuing a Master’s in Software Engineering of Distributed Systems at KTH, where he is currently researching conflict-free replicated data types in the context of edge computing for
his thesis.

The Ultimate 1Million Movies Dataset
Martín Bravo - KTH Student

This serverless ML system predicts IMDb ratings using a dynamic Kaggle dataset. It features four automated pipelines: historical data backfilling, daily feature updates, model training, and inference for new movies. A user-friendly interface allows users to explore movies, view predicted ratings, and compare them with actual IMDb scores.

Speaker Bio:
Martín is a Computer Science and Engineering student from Chile, currently an exchange student at KTH Royal Institute of Technology in Stockholm. He has worked with the National Center for Artificial Intelligence and SoyMomo in Chile and is currently interning at Hopsworks as a Machine Learning Engineer, focusing on LLMs and ML pipelines.

JobsAI
Kolumbus Lindh - KTH Student

Finding the right job among thousands of listings can be overwhelming. JobsAI streamlines this process using vector embeddings and similarity search to match users’ resumes with the most relevant job postings. By leveraging Natural Language Processing (NLP) and machine learning, the platform eliminates inefficiencies associated with traditional job search methods. JobsAI retrieves job listings via Arbetsförmedlingen’s JobStream API and processes user-uploaded resumes to calculate compatibility scores.

Speaker Bio:
Kolumbus is a master's student in Industrial Engineering and Management at KTH, specializing in Machine Learning. With a background in AI consulting and software development, he enjoys finding practical ways to integrate AI into businesses.
–

About the event

Date: February 27th, 17:30 - 20:30
Location: Hopsworks Office (Åsögatan 119, Plan 2, 116 24 Stockholm)
The venue hosting us is the Hopsworks Office. As the office is sometimes difficult to locate we have made this map for everyone to follow.
Directions: 2-minute walk from Medborgarplatsen.
Tickets: Sign up required. Anyone who is not on the list will not get in. The event is free of charge.
Capacity: Space is limited to 60 participants. If you are signed up but unable to attend, please change your RSVP by February 26th.
Food and drinks: Food and drinks will be provided.
Questions: Please contact the meetup organizers.

–

Code of Conduct
The NumFOCUS Code of Conduct applies to this event; please familiarize yourself with it before attending. If you have any questions or concerns regarding the Code of Conduct, please contact the organizers.