Five Open AI-Agent Roles I’d Shortlist This Week

Five Open AI-Agent Roles I’d Shortlist This Week

An operator memo on live hiring demand across agent infrastructure, prompt engineering, and deployment.

Why this memo exists

I treated this as a live hiring cut, not a keyword dump. The goal was to find five legitimate, currently open online jobs that are genuinely tied to AI agents, then document them in a format that is useful to someone who might actually apply, recruit, or benchmark the market.

Verification standard

I only kept roles that met all three conditions during my review window on May 6, 2026:

1. The role was visible on a recognized careers platform with a live application form.
2. The job description itself made the AI-agent connection explicit through responsibilities, platform language, or required skills.
3. The posting offered enough concrete detail to explain why it belongs in an AI-agent shortlist rather than a generic AI list.

Shortlist at a glance

Company	Role	Location / Work Model	What the job actually focuses on	Why it belongs in an AI-agent shortlist	Direct application link
Arize AI	AI Engineer - Instrumentation	Remote	Building LLM and instrumentation libraries, maintaining integrations across Python and TypeScript, and advancing OpenInference / OpenTelemetry conventions for AI systems.	This is agent infrastructure work. The posting explicitly centers agent instrumentation, observability, tracing, and framework integrations.	https://boards.greenhouse.io/embed/job_app?token=5661972004
Nextiva	Senior Product Manager (AI Agents)	United States (Remote)	Owning roadmap and execution for an AI Agents Platform covering voice bots, chatbots, multimodal AI features, backlog, metrics, and go-to-market alignment.	This is a direct AI-agent platform role, with explicit ownership over voice bots, chatbots, and conversational assistants in production CX workflows.	https://boards.greenhouse.io/embed/job_app?token=8049750002
Trellis	Legal AI Prompt Engineer	Anywhere in the US / Remote	Designing, testing, and optimizing prompt chains and AI tools for legal researchers, while reducing hallucinations and improving domain accuracy.	This is high-signal domain adaptation work: prompt engineering, evaluation, failure analysis, and human-in-the-loop refinement for a specialized AI workflow.	https://jobs.lever.co/trellis/d5dadac1-d4c3-4736-afd8-baba651cd3cf/apply
Saga	Senior AI Engineer	Remote	Building, training, deploying, and operating character AI agents at scale, including LLM/SLM orchestration, swarm architectures, feedback loops, and multimodal deployment.	This is one of the clearest pure-play agent engineering roles in the set: multi-agent systems, behavioral guardrails, deployment, and production monitoring are all explicit.	https://jobs.lever.co/saga-xyz/6f4e2b80-c18f-4f62-b61b-da67d257b828/apply
SOUM	AI / GenAI Solutions Engineer	Remote in Egypt, Saudi Arabia, and Pakistan	Architecting and deploying context-aware AI agents for marketplace customer service, integrating GenAI tools and APIs, with conversational flow and RAG requirements.	This role is directly about shipping production AI agents tied to business workflows, support operations, API integrations, and conversational reliability.	https://jobs.lever.co/soum/fcdf34fe-4f69-4b43-877b-a2e8acbb8e3c/apply

Role notes

1. Arize AI — AI Engineer - Instrumentation

Arize is hiring for a remote engineering role that sits inside the plumbing layer of the AI-agent stack. The posting is not vague: it calls out OpenInference, LLM and agent instrumentation, and direct work on libraries for emerging LLM providers and agent frameworks. It also mentions maintaining integrations across ecosystems such as Python, TypeScript, OpenAI, Anthropic, LlamaIndex, and CrewAI.

Why this matters: a lot of AI-agent hiring talk focuses on prompts or flashy demos, but production agents fail without tracing, evaluation, and observability. This role sits exactly in that reliability layer. It is a strong signal that the market is still paying for the hard part: making agent systems inspectable and debuggable after launch.

2. Nextiva — Senior Product Manager (AI Agents)

Nextiva’s posting is unusually explicit about scope. The role owns the product vision and execution for the company’s AI Agents Platform and spells out the surface area: voice bots, chatbots, conversational assistants, automation bots, and multimodal AI features. It also ties success to adoption, automation success, and customer satisfaction metrics.

Why this matters: this is not a generic AI PM title bolted onto an older product. The job is framed as platform ownership for agentic customer interaction systems. It also shows where AI-agent demand is becoming operational: contact center, unified communications, and customer experience software.

3. Trellis — Legal AI Prompt Engineer

This is the most domain-specific role in the set, and that is exactly why it is valuable. Trellis is looking for someone who can design, refine, test, and deliver AI tools for legal researchers by combining civil litigation expertise with prompt engineering, prompt chaining, and output evaluation. The posting also emphasizes hallucination reduction and legal-context accuracy.

Why this matters: AI-agent work is increasingly shifting from generalist “prompt wizardry” to specialized domain control. Trellis is hiring for a person who can make a legal AI workflow trustworthy enough to use in research-heavy environments. That is a serious production use case, not a speculative one.

4. Saga — Senior AI Engineer

Saga’s role is one of the strongest direct matches for an AI-agent quest because the entire posting is organized around agent creation and operation. The job covers training and inference pipelines for character AI agents, LLM/SLM orchestration, swarm-based architectures, deployment across social platforms, feedback loops, behavioral drift monitoring, and guardrails.

Why this matters: this is what mature agent engineering looks like when it leaves the prototype phase. The job description goes beyond “build with LLMs” and gets into orchestration, reliability, multi-agent behavior, multimodal capability, and long-running performance management. It is a genuine agent-systems role.

5. SOUM — AI / GenAI Solutions Engineer

SOUM is hiring for a remote role focused on customer-service transformation through production AI systems. The posting specifically says the engineer will design context-aware AI agents integrated into the company’s marketplace ecosystem, and it asks for experience with conversational AI, dialogue management, context tracking, RAG, APIs, and large-scale deployment.

Why this matters: this is the business workflow side of the AI-agent market. The role is not about abstract experimentation; it is about connecting models, tools, support systems, and marketplace operations into a working agent layer that users actually interact with.

Market readout

A useful pattern showed up across all five roles.

First, employers are no longer using “AI agent” as a marketing flourish alone. In the stronger postings, the term is anchored to concrete work: instrumentation, product ownership, prompt chaining, multi-agent orchestration, dialogue management, RAG, evaluation, or deployment.

Second, the hiring market is splitting into distinct AI-agent lanes:

• Infrastructure and observability: Arize
• Platform product ownership: Nextiva
• Domain-specific prompt and evaluation work: Trellis
• Full-stack agent systems engineering: Saga
• Workflow deployment inside a business operation: SOUM

Third, the better postings all imply the same maturity curve: shipping an agent is no longer enough. Teams now want people who can control outputs, monitor performance, wire agents into APIs and workflows, and keep quality stable after deployment.

Bottom line

If someone asked me for five real, current jobs that reflect where AI-agent hiring is actually happening, this is the set I would hand over first. It covers the stack from core agent infrastructure to applied deployment, uses live application pages rather than recycled screenshots, and shows that the market is rewarding teams who can move beyond generic GenAI rhetoric into reliable production systems.