AI Long-Context Chunking Research Engineer

Lontex Tech

Remote · $160k – $200k / year · Full-time · Posted 4 months ago

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

Context windows are getting longer. 128K, 200K, 1M tokens. The naive conclusion is that chunking becomes less important — just put everything in context. The sophisticated conclusion is that chunking strategy becomes more important, not less: with 1M tokens in context, the model's attention is spread across 800,000 tokens of potentially irrelevant content, and the signal-to-noise ratio of the relevant passage plummets. Long-context retrieval that doesn't chunk intelligently is worse than short-context retrieval that chunks well. We build LLM serving and retrieval infrastructure handling 140 million API calls per month. Long-context retrieval strategy is an active research area for us, and we need a senior research engineer who can push the state of the art for our production use cases. The question we're trying to answer is: for documents too large to fit in any context window, and for queries that require multi-hop reasoning across distributed document sections, what chunking and retrieval strategy minimizes information loss? You'll lead this research program: designing experiments, implementing novel chunking approaches, building evaluation infrastructure, and translating findings that generalize into production retrieval improvements. Expect to publish.

Fully remote.

Responsibilities

Design and implement novel chunking approaches for long-document retrieval with empirical evaluation
Build multi-hop retrieval strategies for queries requiring information synthesis across document sections
Develop information-theoretic chunking evaluation measuring what each strategy preserves versus loses
Implement context window utilization analysis identifying the effective retrieval threshold per model family
Create production retrieval benchmarks comparing long-context versus chunked retrieval strategies
Publish research findings in LLM systems or information retrieval venues

Requirements

6+ years in ML research with retrieval systems, LLM applications, or NLP engineering
Deep understanding of attention mechanisms, context window behavior, and RAG architectures
Strong Python and PyTorch skills with information retrieval evaluation methodology
Research track record with publications or significant contributions in LLM systems or IR
Ability to design rigorous experiments that distinguish genuine improvement from benchmark artifacts