Applied AI Engineer, Document Understanding

AI Chopping Block; San Francisco (CA); On-site; USD 150,000 - 210,000; Full time

Job summary

LlamaIndex in San Francisco is seeking an AI Research Engineer to join our document understanding team. This role sits at the intersection of applied research and strong engineering, working on vision-language models, document processing, data curation, synthetic data generation, benchmarking, training, fine-tuning, and post-training.

The goal is to make our document AI systems more accurate, faster, and more cost-effective in production.

Qualifications

• Hands-on ML with benchmarking and model training experience.
• Ability to translate research into production-ready code.
• Skill in designing experiments and measuring results.

Responsibilities

• Develop and train vision-language models for document processing and understanding.
• Build data pipelines for curation, synthetic data generation, labeling, and benchmarks.
• Evaluate base models and perform post-training or fine-tuning to hit targets.
• Improve model accuracy, latency, and cost in real-world workflows.
• Design and maintain benchmarks for extraction quality, OCR, and reasoning.
• Work with messy real-world documents (PDFs, scans, forms, multi-page).
• Collaborate with engineering to move research prototypes to production.
• Engage with customers to turn requirements into experiments and improvements.
• Stay close to latest research in vision-language models and document AI.
• Use modern AI coding workflows to move quickly.

Skills

Python

PyTorch

Vision-language models

NLP

Benchmarking

Production-ready code

Tools

vLLM

Pydantic

uv

ruff

mypy

Claude Code

Cursor

Job description

LlamaIndex in San Francisco is seeking an AI Research Engineer to join our document understanding team. This role sits at the intersection of applied research and strong engineering, working on vision-language models, document processing, data curation, synthetic data generation, benchmarking, training, fine-tuning, and post-training.

The goal is to make our document AI systems more accurate, faster, and more cost-effective in production.