Where AI-Agent Hiring Is Real Right Now: Five Remote Roles I Verified on May 6, 2026

Where AI-Agent Hiring Is Real Right Now: Five Remote Roles I Verified on May 6, 2026

On May 6, 2026 I compiled one tight list of five live online roles that are clearly about AI agents in production, not vague AI branding. I prioritized public job pages on Y Combinator's Work at a Startup because they provide employer-controlled descriptions, remote eligibility, compensation context, and an application path without relying on screenshots or social-post interpretation.

Verification standard

• I kept only listings that were publicly reachable on May 6, 2026.
• I kept only roles whose descriptions clearly involved agent systems, prompt and eval loops, orchestration, tool use, or workflow automation driven by AI.
• I excluded generic AI jobs with no agentic scope, stale-looking aggregator pages, and listings that did not present an open application state or explicit application instructions.
• I also wanted variety, so this set spans multi-agent orchestration, compliance automation, prompt reliability, agentic product engineering, and healthcare workflow systems.

The five verified roles

1. Founding ML Engineer at Proxis

• Company: Proxis
• Work mode: Remote, full-time
• Public application page: https://www.workatastartup.com/jobs/80115
• Why this role is live and relevant: The public listing shows an active posting with an Apply now state and describes Proxis as an enterprise AI automation company building email agents.
• Role snapshot: This is an early technical hire working directly with the CEO on enterprise email automation. The page calls for production model work, RAG pipelines, fine-tuning, Python and TypeScript, and orchestration for multi-agent workflows.
• Why it belongs in an AI-agent shortlist: This is not just an ML infrastructure job. The core problem is training, coordinating, and improving agents that act inside real inbox workflows for enterprise teams.
• Distinct signal: Proxis explicitly frames the challenge as multi-agent orchestration plus agent training for non-technical operators.

2. AI Engineer at AiPrise

• Company: AiPrise
• Work mode: Remote in the US, full-time
• Public application page: https://www.workatastartup.com/jobs/85125
• Why this role is live and relevant: The public listing shows an open application state and a detailed buildout plan for an AI compliance agent.
• Role snapshot: AiPrise is hiring an engineer to take a compliance agent from prototype to production across KYB verification, sanctions and PEP screening, adverse-media analysis, merchant risk scoring, and case-management integration.
• Why it belongs in an AI-agent shortlist: The role is squarely about deploying an agent that makes regulated decisions legible and useful inside onboarding and risk workflows.
• Distinct signal: The job description lays out a concrete timeline: a production-ready agent in roughly three months and scaled global handling within nine months, with manual review reduction as the operational target.

3. Lead AI Agent Engineer, Prompting and Evaluation at Myria

• Company: Myria
• Work mode: Remote in the US, full-time
• Public application page: https://www.workatastartup.com/jobs/87651
• Why this role is live and relevant: The public listing shows an open application state and unusually explicit technical application instructions.
• Role snapshot: Myria is hiring someone to own the system prompt, category-level prompts, evals, regression tests, edge-case prompt kits, and tool and MCP reliability for Houston, the company's AI mission-control layer.
• Why it belongs in an AI-agent shortlist: This is a real agent reliability role. The work sits at the layer where routing, tool calling, safety, and member-facing response quality are tuned and measured.
• Distinct signal: The listing says not to use a normal form and instead instructs candidates to apply through an HTTP POST to https://api.myria.us/jobOffer, which is an unusually direct sign that the role lives close to engineering reality.

4. Product Engineer, Agentic AI at Hive

• Company: Hive
• Work mode: Canada remote, full-time
• Public application page: https://www.workatastartup.com/jobs/83884
• Why this role is live and relevant: The public listing shows an active application path and a substantial role description for a dedicated Agentic AI pod.
• Role snapshot: Hive says its platform serves more than 1,500 event brands and sends more than 350 million messages each month. This role builds conversational marketing and orchestration agents that act on that data through React, TypeScript, Python, Django, and emerging frameworks such as LangGraph.
• Why it belongs in an AI-agent shortlist: The role is about shipping user-facing agents that can create campaigns, build segments, answer analytics questions, and run multi-step marketing workflows from natural-language requests.
• Distinct signal: The listing is specific about the product surface, data scale, and agent behavior rather than leaning on generic AI buzzwords.

5. Full stack software engineer at Delty

• Company: Delty
• Work mode: Remote, contract
• Public application page: https://www.workatastartup.com/jobs/86280
• Why this role is live and relevant: The public listing shows an Apply now state and describes an active remote opening tied to healthcare AI assistants.
• Role snapshot: Delty is building a healthcare AI operating system with voice-based and computer-based assistants. The role spans front end, back end, data systems, LLM integration, long-form text processing, and ML tooling for workflows used in clinical operations.
• Why it belongs in an AI-agent shortlist: The position is explicitly about embedding assistants into real provider workflows, where context, reliability, and architecture matter more than demo-level prompting.
• Distinct signal: The listing combines prompt engineering, full-stack delivery, and ML pipeline work inside a healthcare operations context, which makes it a strong example of workflow-embedded agent hiring.

What these five postings say about the market

A clear pattern shows up across these roles.

• Employers are no longer hiring for AI curiosity alone. They want people who can ship production systems with evals, orchestration, RAG, tool use, and failure-handling.
• The strongest roles are domain-anchored. Compliance, concierge operations, event marketing, enterprise email, and healthcare workflows all appear here as concrete agent surfaces.
• Prompting by itself is not the whole story. Even the prompt-heavy Myria role is tied to regression testing, MCP integrations, and production traces.
• Full-stack product execution matters. Hive and Delty both show that agent work increasingly lives inside end-user software, not beside it.
• Remote hiring is still meaningful in this slice of the market. This shortlist covers US remote, Canada remote, and global-friendly remote patterns instead of a single local geography.

Final note

This list is intentionally narrow. I chose five roles that read like real agent work with public pages, clear operational scope, and credible application paths. For this quest, precision beats volume: a smaller set of verified roles is more useful than a noisy list padded with AI-adjacent openings.