Research Engineer, Agents

R&D · Full time

About Pelles

Pelles transforms trade contractors' knowledge, documents, and workflows into AI-powered systems — so projects run the way their best people would. We're building agents that read drawings, ground claims in spec paragraphs, and orchestrate real workflows across Procore, Autodesk, and the project document library.

Key Responsibilities

Design and ship agent architectures that solve real construction workflows: RFI drafting, submittal compliance, schedule Q&A, pay-app validation
Build evaluation pipelines and domain-specific test sets from real customer projects
Push the boundaries of long-context retrieval and multi-document grounding
Work directly with customers — superintendents, estimators, project managers — to validate and refine
Own a workflow end-to-end, from research idea to production rollout

Requirements

Strong background in modern ML — LLMs, retrieval, agent systems
Experience shipping AI to production, not just notebooks
Comfort with messy, document-heavy domains
A real interest in evaluation as a discipline, not an afterthought

Bonus

Published work on agents, retrieval, or evaluation
Experience with vision-language models on document or technical-drawing data
Construction or AEC industry exposure

What we offer

Equity in an early-stage AI company at the start of its scale-up
Work on problems that show up on real jobsites the next day
Small team, high agency, no management ladder

To apply, send your CV to [email protected]