AI Researcher

Vetto; United States; Remote; USD 120,000 - 170,000; Full time

Benefits offered by this job

Stock option plan

Job summary

Vetto is seeking a researcher to explore how AI models develop new capabilities by designing experiments from datasets and evaluations through analysis, then turning findings into improved data or training strategies.

Global, remote-first work with periodic in-person gatherings, ownership from research question to conclusion, and collaboration on evaluation, data, and post-training challenges. Stock options are included.

Qualifications

• Comfort writing code, analyzing data, and independently moving experiments forward.
• Ability to design rigorous experiments and draw useful conclusions from model evaluations.

Responsibilities

• Design and run experiments to study model behavior and test research ideas.
• Build datasets, benchmarks, agent environments, rubrics, and evaluation methods for capabilities that are not yet well measured.
• Analyze model failures and use findings to inform data and training approaches.
• Use fine-tuning or post-training experiments when useful to test hypotheses and measure capability changes.
• Work with research partners to turn broad model-development goals into concrete investigations.
• Communicate results through evaluations, technical reports, benchmarks, or papers.

Job description

Role overview

Research how to help AI models develop new capabilities and determine whether training approaches work. You’ll take research questions from experiment design through datasets, evaluations, and analysis, then turn findings into improved data or training strategies. Projects may involve agents, model evaluation, human or synthetic data, failure analysis, or post-training.

Responsibilities

• Design and run experiments to study model behavior and test research ideas.
• Build datasets, benchmarks, agent environments, rubrics, and evaluation methods for capabilities that are not yet well measured.
• Analyze model failures and use findings to inform data and training approaches.
• Use fine-tuning or post-training experiments when useful to test hypotheses and measure capability changes.
• Work with research partners to turn broad model-development goals into concrete investigations.
• Communicate results through evaluations, technical reports, benchmarks, or papers.

Requirements

• Comfort writing code, analyzing data, and independently moving experiments forward.
• Ability to design rigorous experiments and draw useful conclusions from model evaluations.
• Research or engineering experience is useful, though no single background is required.

Benefits and work setup

Global, remote-first work with periodic in-person gatherings every few months. The role offers ownership from research question to conclusion and collaboration on evaluation, data, and post-training challenges. Compensation includes a stock option plan.