AI Agent Operations SpecialistLocation: Remote, must be available during Pacific time business hours.
Duration: 6 months, with possibility of extension based on project needs.
Schedule: Flexible, estimated 5–30 hours/week depending on project demands at any given time.
Pay: $100–$115/hr
About the Project
Our client, a leading AI research organization, runs a research project studying how AI agents behave when they operate autonomously on the open internet, outside the client's internal systems, over long time horizons. The project runs dozens of AI agents that build and maintain open-source software, deploy small web apps, publish content, and interact with real external services and people. Every agent runs inside monitored infrastructure with strict guardrails on what it is allowed to do.
The agents constantly hit steps that only a human can or should do: creating accounts, passing captchas and identity checks, handling credentials, and getting sign-off before using a new service. Today the research engineers do this by hand, and it has become a significant bottleneck. This role exists to own that work.
The Role
You are the human-in-the-loop for the agents. You field their requests, operate a remote browser on their behalf for anything that needs a real person, keep the account and credential inventory clean and secure, and tell the team what you are seeing. You will spend more time interacting with AI agents than with people.
What You'll Do
- Monitor the agent request queue and act on requests promptly during your working hours.
- Drive a remote, in-browser interface to complete signups, logins, captchas, and email/phone verifications for services on an approved list (e.g., code hosting, hosting/CDN providers, domain registrars, cloud providers, and similar).
- Create, store, rotate, and retire passwords, API keys, and 2FA codes in a shared password manager, following handling rules exactly.
- Mint and distribute API keys/access tokens for agents from admin accounts; set spend and permission limits where supported; revoke keys when an agent is retired.
- Maintain an accurate inventory of which agent holds which accounts, domains, and keys, and clean up after experiments end.
- Communicate with the agents to clarify requests, push back on anything out of bounds, and escalate anything new or ambiguous to the research team rather than improvising.
- Help maintain the list of approved and disallowed services/activities as new request types come in.
- Write a short daily summary: what was requested, what you did, what got blocked and why, and anything notable.
What We're Looking For
- Comfortable and curious working with AI agents all day; ideally already a frequent user of AI assistants/agent tools.
- Trustworthy with sensitive credentials — treats passwords, API keys, and recovery codes with real care and follows handling rules without shortcuts.
- Good judgment and discretion — notices when something feels off, stops and asks, and is comfortable telling an agent no.
- Technically fluent power user, not necessarily an engineer. Knows what an API key, a DNS record, a code repository, and a one-time 2FA code are, and can follow a technical setup guide without hand-holding.
- Clear, concise written communication — most reporting is async in chat and shared docs.
- High tolerance for repetitive, detail-heavy work — much of the day is forms, verifications, and chat.
- Reliable, predictable availability during Pacific-friendly business hours, since agents stay blocked until you act.
- Willingness to complete an enhanced background/security screening given the sensitivity of the work.
Nice to Have
- Prior work in trust & safety, IT help desk, QA, research operations, identity verification (KYC), or data-labeling programs.
- Light scripting or spreadsheet automation.
- Experience with password managers, authenticator apps, and remote-desktop or VNC-style tools.
#RTA
#LI-BK1
Never miss an opportunity! Create an alert based on the job you applied for.
