Leads the team this role joins
Decides on early hires
Runs hiring at the company
From your own Gmail and LinkedIn.
About the role
This is a mid-level, full-time remote research engineer role in the United States for someone with at least three years of experience in search or NLP. The engineer evaluates and improves multilingual semantic search for channel search across six or more languages. The work centers on benchmarking relevance, measuring retrieval quality, improving search methods, testing models within latency limits, and documenting results.
What you’ll do
- Build a relevance benchmark for channel search in six or more languages, including query sets and relevance judgments.
- Measure retrieval quality by language and query type using nDCG and recall at k, and analyze failure cases.
- Improve embedding inputs, vector weighting, and the fusion of semantic and lexical results.
- Test stronger embedding models and re-rankers within a CPU-only latency budget.
- Document the method and results.
What they’re looking for
- Experience in information retrieval, search, or NLP engineering.
- Hands-on production experience with dense retrieval and hybrid search.
- Strong Python skills.
- Experience with a vector database such as Qdrant, Milvus, or Weaviate.
- Experience with multilingual embedding models.
- Model export and optimization with ONNX.
- Experience with low-resource languages.
- Clear written English.
Nice to have
- Published information retrieval or NLP work.
- Open-source contributions.
What’s on offer
- Full-time, remote working arrangement in the United States.
Skills
Questions about this role
Is this research role remote?
Yes, it is remote and based in the United States.
What experience is required?
Experience in information retrieval, search, or NLP engineering, with at least three years of experience overall.
Which vector databases are relevant?
Experience with Qdrant, Milvus, Weaviate, or a similar vector database is required.