Software Engineer, Knowledge Engine – Data Infrastructure
ExaExa is looking for a Software Engineer to join the Knowledge Platform team in Zürich. This team is building Exa’s knowledge engine, one of the world’s largest open-domain knowledge graphs with hundreds of billions of evidence-backed claims sourced from the web and other data streams. The goal is to enable agents to answer any factual question with unprecedented precision and completeness, using two halves: construction pipelines that determine trust and retrieval engines that guide searching. The company is aiming to empower teams with a foundation that scales and stays correct as data evolves.
In this role, you’ll focus on the knowledge construction pipelines, turning streams from a very fast web crawl and vendor datasets into a graph that encodes what is true, while tracking provenance and timestamps. It’s greenfield work with a massive scope: there isn’t a ready-made system for modeling the world from open data, so you’ll shape how claims are represented, how trust is assigned, and how the graph grows so downstream work can rely on a solid, evolving foundation.
This is a full-time, onsite position based in Zürich. It’s a fit for someone who enjoys building large-scale data systems, cares deeply about data quality and reliability, and wants to shape core infrastructure that other teams will depend on.
The day-to-day that powers a world-spanning knowledge graph
Your everyday work centers on ingesting and harmonizing data from live web crawls, vendor deliveries, APIs, and other sources into a single typed graph. You’ll implement extraction that converts pages and records into ontology-shaped claims, each with provenance, a timestamp, and a confidence score.
Bringing every source in cleanly matters too: you’ll maintain versioned, append-only histories with schema evolution, correct late-arriving records, and provide replayable cursors so every claim can be traced back to the exact delivery that produced it.
Deciding what to trust is a core part of the job. You’ll perform entity resolution across billions of records to land items into canonical people, companies, places and publications, then run truth discovery that weighs source reliability, recency and corroboration to reconcile conflicting claims.
Keeping the picture current and convergent is essential since streams arrive at different speeds and in various orders. You’ll build incremental computation over change data feeds, ensure each update is a diff, and design everything so jobs are idempotent and resumable. Readers will always see committed, consistent versions, even when runs interleave or fail.
Finally, you’ll measure how well the system is performing: labeled evaluations for extraction and resolution, per-source coverage and correctness metrics, and lineage that shows where things came from and why a certain decision was made.
What you’ll need to bring to the table
You’ve built and operated large-scale distributed data pipelines in production over petabytes of structured and unstructured data, using Spark on a lakehouse format such as Delta Lake, Iceberg or Hudi
You understand incremental processing deeply, including change data capture, idempotent writes, backfills and reprocessing
You’re strong in Python and at least one systems language, and you can read and tune Spark query plans
You prioritise correctness, reliability and data quality as much as you do scale
Experience with entity resolution, record linkage, knowledge graphs or ontologies
Background in information extraction or truth discovery at scale
Familiarity with Lance or other columnar/vector-native formats
Compensation, benefits and practical details
The role offers an annual salary in the range of 180,000.00 USD to 350,000.00 USD. It is a full-time, onsite position located in Zürich, Switzerland. Exa is an equal opportunity employer, and all qualified applicants will be considered without regard to protected characteristics or other legally protected statuses.
Advice for applicants
Lead with a project that shows you moved a billion-row dataset from first delivery to queryable entities in days. In your resume, foreground a concrete example where you built or operated a data pipeline at that scale and explain the end state you achieved for downstream consumers.
When you describe the must-have skills, tie them to specific mechanisms you’ve used: how you implemented change data capture, how you ensured idempotent writes, and how you backfilled or reprocessed data without breaking downstream readers. Mention any metrics or outcomes you can quantify, like how you kept latency or accuracy within target ranges.
For interview prep, be ready to discuss how you design provenance and trust decisions, how you handle out-of-order streams and eventual convergence, and how you would diagnose a fault that caused a source or stage to diverge. You may be asked to walk through a scenario where a late-arriving record changes a previously produced claim and how you would reconcile it without disrupting readers.
When crafting questions to ask, focus on how the team measures precision and completeness of the graph, and how they track lineage to pinpoint where a problem originated. A good question is to ask how they balance freshness with consistency and what the current bottlenecks are in keeping hundreds of billions of claims up to date.