VBT Task Author
Core
Build evidence-based web-research benchmark tasks to expose weaknesses in state-of-the-art AI models by discovering hard-to-surface facts and diagnosing model failure mechanisms.
Role type
Senior IC analytical researcher and red-teaming specialist
Builds
AI evaluation frameworks and benchmark tasks for frontier LLMs
Domain
Artificial Intelligence / Large Language Models / Web Research
Deliverable
production ML models
Required skills
investigative research, fact-checking, AI benchmark task authoring, red-teaming, technical QA review, web search, Open Source Intelligence (OSINT), model trajectory debugging, precision writing
Preferred skills
prompt-injection task design, reasoning-trap task design, advanced OSINT, published investigative research
Technologies
GAIA, BrowseComp, SimpleQA, SWE-bench, Terminal-Bench, Humanity's Last Exam (HLE)
Responsibilities
Discover obscure facts that cannot be surfaced via simple search queries, verify facts against primary sources and document URLs, draft clear benchmark tasks to expose model weaknesses, execute tasks against SOTA models to trigger failures, trace model logs to diagnose root causes of failure, apply validated research patterns across new domains
Seniority
Senior, hands-on IC