Master Thesis Project

Ready to not just envision the future, but to actively build it? At Umain, we’re shaping the next generation of digital products,

and we invite passionate graduate students to join us in crafting groundbreaking solutions that truly make an impact.

With our motto "Shape, Ship, Scale," we empower our team to not only envision the future but also play a pivotal role in building

it. We're a dynamic team where you can shape the future, ship groundbreaking solutions, and scale your abilities in many different

ways. We're on the lookout for passionate graduate students to join us, bringing enthusiasm, creativity, and discovery along with

them. Bring your ideas and we’ll make it happen together.

• A fun and supportive work environment where you have the opportunity to develop your skills through unique challenges and

fantastic colleagues

We believe in nurturing talent, which is why we've established a supportive learning environment. Your growth and development are

important to us. With guidance from experienced team members, you will build your skills and gain valuable industry experience.

At Umain, diversity is our strength. We understand that life doesn't pause when you walk into the office. That's why we've

cultivated an environment that supports a healthy work-life balance. We welcome interns from all walks of life, recognizing the

richness that different backgrounds bring to our team.

Technology is always evolving, bringing new challenges and opportunities for collaboration and learning. Collaborating with us,

you'll work alongside developers, designers, analysts, testers and more to observe how these challenges are tackled, sharing in

the knowledge and growing together.

We want to make sure you leave your master’s thesis with practical skills and insights that support the next steps in your career.

At the same time, we want your thesis work to create real value, contributing insights and results that make a difference in the

projects and challenges our teams work on every day.

If you're a budding graduate student who thrives on innovation, values a supportive and inclusive workplace, and is eager to learn

and contribute, we'd love to hear from you. Let's build something awesome together.

Umain is a digital product consultancy. We shape, ship and scale digital products at speed. Founded in Stockholm in 2001, we are a

growing team of 100+ experts in digital product strategy, design, and technology.

We are a part of Eidra, a group of top-notch companies specializing in strategy, creativity, and engineering. As a collective we

guide leaders in making meaningful and lasting change.

We do not offer a list of pre-defined projects. Instead, we encourage you to take an active role in envisioning future digital

products. Kindly submit your proposals, ideas or topics and we will assess potential alignments.

Umain’s work is to produce high quality digital products for our customers, and the primary platforms for this are therefore web

and mobile. But we work in a variety of fields and concern numerous topics, so any thesis area concerning the topics below are of

interest to us:

• AI-assisted software delivery. How coding agents change team structure, estimation, code review, quality and junior-developer

growth.

• Building reliable products on top of LLMs. Evaluation, testing, guardrails, cost and latency, and what "done" means when the

behaviour is probabilistic.

• Agentic systems in real business workflows. Where autonomy pays off, where humans need to stay in the loop, and how to design

review and trust.

• AI-native UX. Interfaces that go beyond the chat box, such as adaptive, generative or conversational interaction. This includes

how users build trust in, and calibrate against, AI output.

• Messy-data pipelines. Turning inconsistent real-world inputs like spreadsheets, PDFs and free text into something products can

rely on.

• Regulation and responsible tech. The EU AI Act, Cyber Resilience Act, GDPR and accessibility law (the European Accessibility

Act), and how they translate into engineering practice rather than just compliance paperwork.

• Sustainable and efficient software. Measuring and reducing the energy and carbon footprint of apps, cloud workloads and AI

usage.

• Cross-platform and on-device development. The trade-offs between native, Flutter, React Native and web. This includes running

models on-device for privacy and responsiveness.

• Developer experience and platform engineering. What makes consulting teams fast and consistent across many different client

contexts.

• Software for connected products. Automotive, IoT and in-vehicle or companion-app experiences, where hardware, cloud and mobile

meet.

• Consulting in an AI era. How AI changes the way software gets scoped, sold, priced and delivered, and what clients will expect

from a consultancy in the future.

Command Query Responsibility Segregation (CQRS) is an architectural pattern that separates read operations (queries) from write

operations (commands) in a software system into separate paths. Using a ticket booking system as the test case, a system with

heavy read traffic and occasional spikes in bookings, this project measured how CQRS compares to CRUD (Create, Read, Update,

Delete) on speed, capacity, and resource usage.

Designed to address the high resource consumption associated with traditional JVM platform threads, Kotlin Coroutines and Project

Loom’s Virtual Threads offer modern approaches to concurrency. This study directly compared their performance across I/O-bound and

CPU-bound scenarios, assessing key metrics such as memory overhead, CPU consumption, latency, and overall throughput.

“How effective are different strategies for reducing hallucinations in RAG systems applied to internal technical documentation,

and to what extent can long-context prompting replace or supplement RAG for the same purpose?”

This study evaluated methods for minimizing hallucinations in technical documentation systems, analyzing performance metrics such

as accuracy, latency, and cost while comparing optimal Retrieval-Augmented Generation (RAG) configurations directly with

long-context prompting techniques.