Senior AI Research Engineer — Multimodal Systems

Erik Hoffmann

Full-time · Senior · New_York

This role is no longer accepting applications

Similar open roles you can apply to now:

About the role

We are a company built on research. Three years ago we spun out of a university computer vision lab with a specific capability: multimodal document understanding — systems that can reason jointly over text, tables, figures, and diagrams in complex technical and scientific documents. We have two commercial products built on this capability: a patent analysis tool used by IP law firms and a scientific literature review assistant used by pharmaceutical R&D teams. We are Series A, profitable on our two existing products, and planning our Series B for next year. Our research team of eight has produced six peer-reviewed publications since we incorporated, and we intend to keep publishing — not because it is required, but because the research community is where the ideas come from. We are looking for a Senior AI Research Engineer who can operate at the boundary between research and engineering: someone who reads papers and implements them, who designs experiments rigorously and documents them reproducibly, and who cares about the difference between a result that holds up and one that looks good in a demo.

Responsibilities

Design and implement multimodal document understanding models for patent and scientific literature domains
Conduct rigorous ablation studies and comparative evaluations against published baselines
Contribute to research publications — writing, experiment design, and peer review of team members\' work
Work with the engineering team to translate validated research findings into production features in our commercial products
Mentor junior researchers and contribute to our hiring process for research positions

Requirements

5+ years of ML research engineering with a demonstrable track record in multimodal learning, document understanding, or vision-language systems
PyTorch for implementing research architectures from scratch — you read the paper, you understand the math, you write the training loop
Hugging Face for fine-tuning and evaluating vision-language models on domain-specific document datasets
RAG pipeline design for multimodal retrieval — text, figure, and table retrieval from complex scientific and patent documents
Python at a research engineering standard: reproducible experiments, clean implementation, and the discipline to write code that a collaborator can extend
Statistics for rigorous experimental evaluation — effect sizes, ablation design, and the ability to defend methodology in peer review
Publications, preprints, or equivalent public technical contributions are a significant plus

Benefits

$152,000 – $192,000 base salary + equity (Series A, strong trajectory)
Full remote — our research team is distributed across the US and Europe
$3,500 annual research and conference budget
Publication support — time and resources to write, submit, and present significant findings
Access to proprietary patent and scientific literature datasets that are genuinely unusual to work with commercially