Five AI Agent Roles Open Right Now, From Prompt Design to Agent Evaluation

Five AI Agent Roles Open Right Now, From Prompt Design to Agent Evaluation

If you want a clean signal on where AI-agent hiring is real, the best place to look is not generic repost spam. It is the live application page itself.

I screened current listings on May 6, 2026 and kept only roles that met four standards:

1. The application page was live and directly accessible.
2. The job body made AI agents or agentic systems part of the actual work, not just company marketing.
3. The posting was remote or explicitly online-accessible through a current hiring page.
4. The source was a verified company-hosted board or official application page, not a scraped mirror.

This produced a tighter list than the usual "AI jobs" roundup. These five roles cover five different parts of the agent stack: reasoning and guardrails, backend runtime, prompt quality, product ownership, and evaluation infrastructure.

At-a-glance list

1. Airtable — AI Agent Architect, Customer Experience

What the role actually does

Airtable is hiring an architect-level operator to own the technical foundation of its AI-native customer support stack. The listing is unusually explicit about the job surface area: this person is responsible for how support agents reason, retrieve, decide, and act. The page calls out retrieval accuracy, automated resolution rates, guardrails, observability, prompt architecture, and integrations with external systems like billing platforms, CRMs, internal tools, and Airtable APIs.

Why this belongs on an AI-agent list

This is not a generic support-ops role with AI garnish. It sits directly in the classic agent loop:

• retrieve the right context
• decide what action is safe
• execute through tools or APIs
• observe failures and improve performance

The listing even names the failure modes serious agent teams care about: hallucination rates, prompt injection, unintended behavior, and week-over-week agent quality improvement.

Best-fit candidate signal

A strong fit here is someone who has already touched production RAG, prompt versioning, agent guardrails, and systems integration, even if they are not a full-time ML researcher.

2. Cresta — Senior Software Engineer, Backend (AI Agent)

What the role actually does

Cresta is hiring a senior backend engineer to make sure its AI agents are supported by reliable, scalable server infrastructure. The job description centers on backend architectures for AI agent solutions and proprietary models, API design, high-volume interaction handling, cloud performance, security, and cost control.

Why this belongs on an AI-agent list

A lot of agent hiring chatter focuses on demos and prompts. This role is a reminder that production agents break on boring things first: latency, orchestration bottlenecks, weak APIs, brittle services, and poor database performance. Cresta explicitly wants someone who can support real-world agent deployments at scale, not just experiment in notebooks.

Best-fit candidate signal

This is the posting I would send to a backend engineer who already understands distributed systems and now wants to move deeper into agent runtime and production infrastructure.

3. Netomi — Prompt Engineer

What the role actually does

Netomi describes itself as an agentic AI platform for enterprise customer experience, and the role itself is tightly scoped around prompt quality. The Prompt Engineer is expected to craft, optimize, evaluate, and benchmark prompts, collaborate with Customer Success and Data Science, and define tool descriptions for agentic frameworks.

Why this belongs on an AI-agent list

This is a credible example of prompt engineering that is actually agent work. The listing does not stop at "write good prompts." It calls for:

• client-specific prompt design
• tool descriptions for agentic frameworks
• automated testing
• evaluation frameworks
• model benchmarking

That means the role sits close to real deployment quality, not just creative prompting.

Best-fit candidate signal

A strong applicant here would likely be comfortable with prompt iteration, LLM evals, customer-specific business rules, and scripting enough automation to test changes rather than eyeballing outputs manually.

4. Wizard — Senior Product Manager, Agentic AI Experiences

What the role actually does

Wizard positions itself as an AI shopping agent, and this PM role owns how that agent behaves across mobile, web, and messaging. The posting says the PM will define how the agent understands intent, takes action, reasons about context, and supports end-to-end shopping flows. It also mentions work with inference pipelines, agent planning, retrieval, orchestration logic, multimodal interactions, and error-recovery patterns.

Why this belongs on an AI-agent list

This is the product side of agentic systems, and it is serious product work. The company is not hiring a generic consumer PM; it wants someone who can turn ambiguous user needs into structured agent behaviors and partner closely with engineering on planning and orchestration. That is exactly where many agent products win or fail.

Best-fit candidate signal

This is a strong opening for a PM who can translate LLM and orchestration concepts into concrete product decisions, metrics, and shipping priorities without getting lost in hype language.

5. Slingshot Aerospace — Senior AI Engineer, Agentic Evaluation & V&V

What the role actually does

Slingshot is hiring for one of the most technically specific agent roles in this set: evaluation and verification for mission-critical autonomous systems. The listing says the engineer will build and scale evaluation frameworks, benchmarks, and simulation-backed validation systems for multi-step, tool-using, and autonomous decision-making workflows powered by LLMs and reinforcement learning.

Why this belongs on an AI-agent list

This is not an "AI engineer" title stretched to fit the trend. The job is explicitly about validating agentic behavior in high-stakes environments. It covers benchmark scenarios, scoring logic, experiment harnesses, failure analysis, regression detection, SDK interfaces, and even familiarity with orchestration frameworks like LangGraph.

Best-fit candidate signal

If someone understands that the hard part of agents is not only generation but also evaluation under realistic conditions, this is the role in the list that most clearly rewards that mindset.

Why these five are stronger than a generic roundup

The point of this list is not just that the titles contain the word "AI" or "agent." It is that each role sits on a recognizably important layer of the modern agent stack:

• Airtable: retrieval, guardrails, safe actioning, observability
• Cresta: backend runtime, scale, APIs, reliability
• Netomi: prompt design, tool descriptions, benchmarking
• Wizard: product behavior, planning, orchestration, user-facing agent experience
• Slingshot Aerospace: evaluation, V&V, autonomous workflow testing

That spread matters. It shows that the hiring market around AI agents is no longer just asking for one mythical "AI agent builder." Companies are carving the work into distinct functions: architecture, runtime, product, evaluation, and prompt quality.

Final take

If I had to summarize the market signal from these five listings in one sentence, it would be this: the real AI-agent hiring wave is moving from demos to operating systems.

The strongest openings are no longer asking only for prompt fluency. They want people who can make agents retrieve correctly, call tools safely, survive production scale, behave well inside a product, and stand up to evaluation.

That is why these five made the cut.