The role
- COMP
- $175K - $275K
- EQUITY
- Highly competitive equity
- LOCATION
- San Francisco
- WORKPLACE
- On-site
- EXPERIENCE
- 4+ years
- VISA
- None, Visa transfers
- STACK
- Python, CI/CD
- INDUSTRY
- AI, Data, Education, Creator Economy
/roles — ROLE_485
AI Engineer (Backend)
Early-stage AI data company building evals, retrieval, and crawling infrastructure to improve design quality in AI output
The company
An early-stage startup building the data, evaluations, and tooling that help AI models develop better aesthetic judgment, with design as its first domain, for leading AI labs and product companies.
- STAGE
- Series A-stage
- FUNDING
- $10M+ raised
- TEAM
- ~20 people
- FOUNDED
- 2024
JD — the work
About the role
A backend-focused AI engineering role on a small, early team working to improve AI output in subjective areas, with design as the first focus. You would build the backend for agents, including harnesses and memory, along with evaluation, data collection, search and indexing, crawling, and security. The team wants someone with strong opinions about code quality, who uses AI heavily in their own workflow but has no patience for sloppy output, and who reaches for the simplest clear design rather than a complicated one.
What you'll do
- Develop agent harnesses and memory, along with loops that let agents improve themselves
- Generate synthetic data and build the pipelines that evaluate model output
- Run embedding and retrieval services that hold up under millions of calls
- Collect visual content at web scale with crawlers and scrapers
- Serve models and expose APIs for customer-facing products
- Handle security sweeps and map out vulnerabilities
- Improve internal tools and infrastructure so the whole system stays quick and dependable
What they're looking for
- 4+ years of engineering and strong, well-reasoned views on what good code looks like
- Hands-on experience shipping AI-centric products, from agents and systems to models
- An AI-heavy personal workflow paired with low tolerance for sloppy output
- Curiosity about subjective quality and how to make it measurable and verifiable
- Startup instincts: quick to adapt, happy to own work outside any job description, drawn to hard, ambiguous problems
- A bias toward simple, clear solutions over complicated ones
Nice to have
- Open-source or personal projects driven purely by curiosity
- Background at a company known for web crawling, search indexing, or AI training data