Current SOTA models fail at basic Computer Use and agentic workflows because they lack the necessary data scaffolding for simple OS/Browser tasks. Job Title: Member of Technical Staff (Part-Time / Spring) Company: Archal Labs Location: Remote / US-Based Commitment: 20-25 hours/week Archal Labs.
Numbers & Facts
Location
San Francisco, California (Remote)
Description
Job Title: Member of Technical Staff (Part-Time / Spring)
Company: Archal Labs
Location: Remote / US-Based
Commitment: 20-25 hours/week
Archal Labs
Current SOTA models fail at basic Computer Use and agentic workflows because they lack the necessary data scaffolding for simple OS/Browser tasks. They hallucinate actions because they haven't seen enough high-quality, repaired traces of human workflows. We're solving this problem by building the data infrastructure for the next generation of Action Models. We are looking for a researcher to join us this Spring.
What You’ll Build
Dynamic Benchmarking (Arch Engine): Existing benchmarks (OSWorld, WebArena) are static and gameable. We want to engineer benchmarks that are dynamic and reflective of real-world tasks.
Data Scaffolding & SOTA Strategy: In general, doing research into increasing the value of each trace.
Research & Validation: You will assist in writing technical whitepapers and validation studies.
Who You Are
We're mainly looking for PhD students but cracked undergraduates are fine.
You've ideally worked in technical roles at Surge/Scale/Mercor, but this isn't required. Papers on computer use or RL are cool.
You must be a US Citizen. We can't sponsor F-1/CPT/OPT for this role.