About Staffbase

We inspire people to achieve great things together. Our mission is to help organizations unlock the power of inspirational communication with the first AI-native Employee Experience Platform. Our industry-leading and award-winning agentic AI communications channels - intranet, employee app and email solutions - create engaging experiences that connect and empower employees.

Headquartered in Chemnitz, Germany and New York City, with offices in Berlin, London, Sydney, Tokyo, Prague, and Minneapolis–St. Paul, our diverse team of 550+ employees supports 1,500+ customers—reaching over 14 million employees—in transforming their employee experience.

We are proud to be a Unicorn company—privately valued at over $1 billion—demonstrating strong growth, innovation, and lasting impact in our industry. Together, we’re shaping the future of workplace communication.

As a Staff Engineer on the Search team, you'll take on one of our most important platform challenges: making every employee able to find what they need, and get a trustworthy answer, across Staffbase and the tools around it. Today search spans people, posts, pages, files and connected integrations, and it increasingly feeds AI-generated answers. You'll define how retrieval, ranking and grounded answers work at that scale.

Your north star is a search platform that finds the right thing, fast, for the right person. It must respect permissions on every query, work in every language our customers use, and give answers that cite their sources. You'll build the core, set the quality bar, and decide where we use proven techniques and where we experiment, all on a mature, multi-tenant platform.

This is a role for someone who wants to drive, not just deliver. You'll shape the direction of an open-ended problem, and you'll measure your work with real relevance data, not gut feeling.

Our tech stack includes Kotlin, Java, OpenSearch/Elasticsearch, MongoDB, React, TypeScript, Terraform, Docker and Playwright, plus LLM and embedding services for retrieval, reranking and summarisation.

The Team

The Search team sits at the heart of how employees find information at Staffbase. We own the full path from query to result: indexing and embeddings, hybrid keyword and semantic retrieval, ranking and reranking, typeahead, and the AI summaries that answer questions with citations. Search is also a core source of knowledge for Navigator, our AI assistant, and the main way it grounds its answers in company content.

That means two big things. First, we make results relevant: understanding the query, retrieving the right candidates, ranking them well, and doing it in the user's language. Second, we make search safe and fast: every result respects the user's access rights, every tenant stays isolated, and the fast path stays fast. We don't ship relevance changes on a hunch. We run offline evaluations, golden query sets and online signals, and we change one thing at a time.

Our environment

Working with skilled and similar-minded professionals in a team that values diversity and inclusivity
We aim to create a flexible and agile work environment that promotes the well-being and work-life balance of our employees
We are enthusiastic about technology and committed to using it to improve the world
We follow the "You build it, you run it" approach and take pride in it

What you'll be doing

Define and drive the search and retrieval strategy: hybrid (BM25 + vector) retrieval, reranking, query understanding and rewriting, chunking and embedding choices
Build and evolve the core ingestion and query pipeline, from sync and indexing to ranked results and AI summaries with citations
Own search quality: build evaluation tooling and golden sets (including unanswerable and permission-varying queries), define relevance metrics, and turn them into decisions
Protect the non-negotiables at scale: permission-aware retrieval, tenant isolation, multilingual handling, and graceful degradation when a dependency is slow or down
Balance latency, quality and cost, for example what runs on the fast path versus what is allowed on the slow path