The posting
Klarna, briefly
At Klarna, we’re building an everyday finance network, helping over 120 million consumers across 26 countries save time and money, and worry less about their finances. Working here means taking on problems most companies never get to solve, and being hands-on enough that the interesting part of the work lands with you, not someone else — you’ll build with AI, not watch it happen.
This is the stretch zone. Come find out what you’re capable of.
About the role
Klarna’s product catalog covers more than 150 million products, matched to over a billion merchant listings in the US and across Europe. A catalog this big only works if the data is complete and accurate. Wrong match, wrong color, missing attribute, and someone buys the wrong thing or never finds the right one. You’ll join the data science team that prevents that and owns the ML and AI layer of the catalog, so our users and any shopping agent can find the right product at the right price. The catalog powers product discovery, price comparison and agentic commerce.
The work covers three areas:
- Matching: deciding which merchant listings are the same product, and catching wrong matches and duplicates.
- Enrichment: extracting and verifying category, brand, color and other attributes from text and images.
- Quality agents: AI agents that find and fix catalog errors using tools, with access to Hive, our product catalog system.
We’re also building a multimodal foundation model that will support multiple use cases through unified predictions, replacing many specialized models while keeping the speed and cost-efficiency that high-volume decisions need.
It’s a small, high-ownership team that ships to production often, building an AI-native catalog where models match, verify and fix products, with humans only where they add value.
What you’ll do
- Take models from training to production, owning data pipelines, training, evaluation and the serving path.
- Decide which model runs where and how it scales on GPUs, setting latency and cost targets for each.
- Build an evaluation for every change: precision and recall, threshold calibration, labeling and QA design, and checks that offline results hold up online.
- Develop our quality agents and measure how well they fix catalog data.
- Monitor each change after it ships, and explain model decisions to backend, operations and product colleagues.
Who you are
- You’ve taken ML models from training to production in a live system.
- You have hands-on deep learning experience.
- You write production-quality Python: tests, code review, CI/CD, containers.
- You’re rigorous about measuring whether models actually work, and know how to choose metrics and thresholds.
- You’ve worked with LLMs and can reason about quality, cost and latency trade-offs.
- You communicate clearly with colleagues in backend, operations and product.
Bonus points if
- You’ve optimized GPU inference (Triton, TensorRT or similar) or trained deep learning models at scale.
- You’ve worked with multimodal models that combine images and text, including pretraining or fine-tuning, embeddings or similarity search.
- You’ve built LLM agents with tool use through MCP, pydantic-ai or a similar framework.
- You’ve built an ML pipeline from scratch, choosing the orchestration, storage and training stack yourself.
Things you should know before applying
- You’ll be based in Stockholm, working in a team of six data scientists and ML engineers who collaborate with colleagues across Europe.
- GPU capacity can be tight at times, and the team regularly fields questions from colleagues about model decisions.
Please include a CV in English. Concrete beats comprehensive — what you built, what it did, what it cost.Curious to learn more about Klarna and what it’s like to work here? Explore our career site!
