AI Engineer, Retrieval & Quality

Redpine; Stockholms kommun; On-site; SEK 600,000 - 800,000; Full time

Benefits offered by this job

Competitive salary

Meaningful equity

Relocation support

Job summary

Redpine in Stockholm is looking for a candidate skilled in designing and operating retrieval systems that serve licensed data to AI models. This role involves building evaluation infrastructures to ensure quality, translating research into practical applications, and defining metrics that prove system effectiveness. The ideal candidate has production experience with retrieval systems and strong opinions on evaluation methods, alongside proficiency in Go, Rust, and Python. Competitive salary and relocation support offered.

Qualifications

• Experience with search, ranking, recommendation systems, or RAG at scale.
• Knowledge of designing test sets and running A/B experiments.
• Familiarity with embeddings, vector and lexical indexes.
• Ability to differentiate between academic methods and those that work in production.

Responsibilities

• Design and operate retrieval pipelines serving licensed data.
• Build evaluation infrastructure to ensure quality.
• Translate research into production systems.
• Define quality metrics and build dashboards to prove them.

Skills

Production experience with retrieval systems

Strong opinions on evaluation

Comfort with the full retrieval stack

Shipped production systems

Adaptability to evolving problems

Tools

Go

Rust

Python

Job description

The Role

Among other things, you'll design and own the systems that decide which licensed data reaches our customers' models, and the evaluation infrastructure that proves it's better than the alternatives.

In this role, you will:

• Design and operate the retrieval pipeline that serves licensed data to AI agents and models in real time: hybrid search, ranking, fusion, re-ranking
• Build the eval harness that runs continuously, gates deploys on quality, and surfaces regressions before customers feel them
• Translate research from our data science team into production systems, then own them end to end
• Define what "good" means for Redpine's output and build the dashboards that prove it across customers, verticals, and competitors
• Work across the stack as quality demands it, from ingestion signals through embedding pipelines to the API response

The Team

We're a small, fast growing team based in Stockholm, working across infrastructure, data, and AI. The same people designing systems are building, shipping, and running them. Feedback loops are short and ownership is direct.

You’ll work closely with our data scientists, who own research, methodology, and data quality. You’ll own production retrieval and the eval infrastructure that runs on top of that.

We're early enough that many of the core systems are still being defined. That means you won't be scaling someone else's architecture.

What we're looking for

• Production experience with retrieval systems: search, ranking, recommendation systems, or RAG at scale, not just notebooks
• Strong opinions on evaluation, both offline and online. You've designed test sets, run A/B experiments, and argued with data scientists and product managers about whether a metric reflects reality
• Comfortable with the full retrieval stack: embeddings, vector and lexical indexes, learning-to-rank, hybrid systems & knowledge graphs
• Have shipped, not only researched. You know the difference between a method that works in a paper and one that works in production
• Comfort operating where direction evolves quickly and problems aren't yet well-defined
• We work primarily in Go, Rust and Python in a multi cloud environment. We hire for judgment, not stack.

Why this role

Most retrieval and AI engineering roles are about plugging the latest models into an existing product. This isn't.

You’ll be working on questions like:

• How do you prove that licensed data produces better agent outputs than scraped data, with rigor that holds up to a customer's evaluation team?
• What does a retrieval system look like when freshness, provenance, and rights are first‑class signals alongside relevance?
• How do you build evals that are honest enough to fail your own deploys and convincing enough to win customers?

This role sits at the core of whether Redpine is differentiated infrastructure or a nicer API on commodity vector search.

Practical details

• This role is based in Stockholm and requires a valid Swedish work permit or similar eligibility
• Relocation support available
• Competitive salary and meaningful equity