Member of Technical Staff, Alignment

The role

You will run experiments on how environments shape behavior and build tools to monitor agents. You will also research oversight, control, or long-horizon autonomy and publish your findings.

What you will do

• Run experiments on how environments shape behavior
• Break graders to find ways to score well without doing the work
• Build tools that watch agents including monitors, red team harnesses, and safety benchmarks
• Keep environments sealed through network isolation and credential management
• Choose, answer, and publish a research question about oversight, control, or long-horizon autonomy
• Decide what the company will not build and inform partners of safety limits

Required qualifications

• Engineer first who builds test pipelines and learns codebases quickly
• Know LLM safety in depth including attacks, red teaming, monitoring and control evaluations, reward hacking
• Ability to turn a vague worry about a model into an experiment
• Ability to read transcripts closely
• Experience putting research into production systems, especially post-training, distillation, or evaluation
• Write clearly enough that results change what other people do
• Ability to decide fast with incomplete evidence and defend the call with data

Nice to have

• Published work on control, dangerous capability evaluations, oversight, or interpretability
• RLHF or RLAIF experience
• Built or ran a public benchmark or agent task suite
• Experience running many agents at once using sandboxes, containers, or logging
• Advised on AI safety or governance policy

Benefits

• Health
• Dental
• Vision
• Flexible PTO

About the company

Abundant is an applied research lab focused on scaling reinforcement learning for safe and reliable agentic capabilities. The team includes researchers, roboticists, founders, and operators whose work includes the Waymo Driver.