Document AI Benchmark Engineer & Technical Writer

LlamaIndex, Inc.; San Francisco, Northern (CA, KY); Hybrid; USD 180,000 - 240,000; Full time

Job summary

LlamaIndex, Inc. in San Francisco is seeking a highly technical ML engineer to produce compelling technical content while building benchmarks and publishing analyses.

You will work on document AI, run experiments, and translate results into content that drives developer awareness and adoption. This role emphasizes shipping real code and benchmarks over traditional marketing, with autonomy and impact.

Qualifications

• Experience in ML engineering with production Python.
• Ability to read and synthesize research papers quickly.
• Proven track record of high-velocity technical output.
• Scrappy, self-directed with end-to-end execution.
• Strong writing skills for technical content.

Responsibilities

• Design, build, and maintain benchmarks for document parsing and understanding.
• Publish high-quality technical content weekly (blogs, benchmarks, tutorials).
• Stay current with document AI landscape: models, papers, competitors.
• Run experiments and translate findings into publishable artifacts quickly.
• Produce analyses showing capabilities vs. alternatives.
• Contribute to open-source examples, notebooks, and documentation.
• Collaborate with ML team to surface improvements and highlight capabilities.
• Engage with developer community through technical content.

Skills

Python

ML engineering

Writing

Benchmarking

Reading papers

Self-directed

Fast-paced output

Tools

OCR

Layout analysis

Table extraction

Document understanding

Benchmarks

Job description

LlamaIndex, Inc. in San Francisco is seeking a highly technical ML engineer to produce compelling technical content while building benchmarks and publishing analyses.

You will work on document AI, run experiments, and translate results into content that drives developer awareness and adoption. This role emphasizes shipping real code and benchmarks over traditional marketing, with autonomy and impact.