7 hours ago
Zürich, SwitzerlandMid Level
Responsibilities
- Improve web-page parsing for pages that are currently handled poorly and demonstrate the resulting improvements.
- Train models to assess page quality and develop shared definitions and supervision for quality.
- Model credibility, misinformation, source reliability, and whether content is written for readers or crawlers.
- Determine whether documents are semantically identical or meaningfully different for web deduplication.
- Build classification and extraction systems that are accurate and inexpensive enough to run across the web.
- Design supervision for problems without existing labels and evaluate whether they are learnable.
- Trace poor search results to the page-level predictions responsible and correct them at the source.
Requirements
- Master’s or PhD with at least 2 years of relevant experience, or an exceptionally strong undergraduate background.
- Ability to build a transformer from scratch in PyTorch.
- Experience training models that are inexpensive enough to run broadly.
- Comfort building large-scale datasets and working deeply with data.
- Ability to define ground truth for problems where it does not yet exist.
- Strong interest in finding high-quality knowledge and its importance to society.
Tech Stack
Categories
About Exa
Exa builds a web-scale search engine and API for AI agents and developer applications, combining its own crawler, embedding models, and high-performance vector search. It sells usage-based APIs and enterprise integrations that power retrieval-augmented generation, browsing, and automation on live web content. The company runs large crawling infrastructure and dedicated GPU clusters to continuously index and embed pages, and ships low-latency search components written in Rust.