research
Posted Jan 24Early Career Research Engineer
at Parallel
On-site
Requirements
- We've raised $230 million from Kleiner Perkins, Sequoia, Index Ventures, Spark Capital, Khosla Ventures, First Round, and Terrain to build the web for AIs.
- You've worked on information retrieval systems, embedding models, or neural ranking at scale, or you're deeply curious about the fundamental problems that emerge when training models to understand and serve billions of web documents.
- You're comfortable reading papers from SIGIR and RecSys one day and debugging distributed training pipelines the next.
- THE ROLE You'll design and train the models that power Parallel's APIs: the intelligence layer that helps AI agents find exactly what they need from the open web.
- Unlike traditional search engines built for human queries, you're building for AI agents that issue complex, multi-hop queries and expect structured, programmatic responses.
- This is information retrieval reimagined for the LLM era, work that combines classical IR techniques with modern deep learning, applied at a scale that demands new solutions.
Benefits
- Our products are used by leading businesses in sales, marketing, insurance, and coding to build best-in-class AI agents with flexible and powerful programmatic access to the web.
- We're currently valued at $2 billion and we're forming a world-class team of engineers, designers, marketers, sellers, researchers, and operational experts to achieve our mission.
- But succeed an unfair amount. COMPENSATION &
- BENEFITS - Competitive salary - Generous equity - Visa sponsorships - 401K plans - Daily lunch & office snacks - Dinner at the office - Unlimited vacation - Caltrain pass reimbursement
Additional details
- ABOUT US Parallel is a web infrastructure company.
- ABOUT YOU You're a researcher who thinks like an engineer, or an engineer who thinks like a researcher.
- You thrive in the space between theory and production, where elegant solutions must also run efficiently on real infrastructure.
- This means tackling research problems that most labs encounter only at hyperscale: How do you train embedding models that capture semantic intent across diverse query types?