AI Agent DesignerAI startup · Full-time · Nangang Software Park
This startup is building an "AI sales rep for merchants," serving U.S. e-commerce sellers — every design decision you make directly affects those merchants' conversion rates. Most AI agents are judged by how the demo feels; this team is different — they validate with real numbers against a control group. There is no ready-made agent framework waiting for you to fill in the blanks: from observing real buyer behavior, research, and prompt design to implementation and post-launch impact validation, you own the whole path yourself.
Job Requirements at a Glance
- Salary & compensation
- Salary negotiable (regular monthly pay of NT$40,000 or more)Actual pay is determined by your experience and interview results
- Nature of the role
- Full-time
- Location
- Nangang District, Taipei (Nangang Software Park, about 430 meters from MRT Nangang Exhibition Center Station)Within walking distance of MRT Nangang Exhibition Center Station
- Job category
- AI engineers, product managers
* The above terms are current plans; actual compensation and working conditions will be as confirmed by both parties after the interview.
Job description
Core skills for building AI agents end to end
- Start by understanding what merchants and buyers really need
- All the way through design and implementation
- and post-launch impact validation
Deliver fast with AI tools
- Heavy use of Claude Code
- Turn complex agent logic into a usable product within hours
Refine the experience from multiple perspectives
- The merchant's perspective
- The buyer's perspective
- The business outcome perspective — the same design decision often has three different right answers from these three perspectives
Make results measurable
- Iterate based on the actual results of control groups, not gut feeling
Requirements
- Claude Code is your main way of building — not an occasional helper, but your default tool for turning ideas into usable products
- Hands-on experience building or optimizing AI agents (sales agents, customer success agents, or any autonomous agent counts)
- Familiar with one or more of prompt engineering, RAG, agent architecture, tool use and multi-agent collaboration
- A real record of getting things done — not necessarily coding; building a business, growing a highly active community, or designing a system that makes things run smoothly all count
- Precise expression: designing agents is essentially thinking in language—vague instructions make for a vague agent
- You want real ownership and want to build this product as if it were your own business
Why this opportunity
Not a fill-in-the-blanks job
There is no ready-made agent framework waiting for you to fill in. From observing real buyer behavior to validating in production, you walk the whole path yourself.
Validated by numbers, not gut feeling
Results are judged by what the control group actually shows. Whether it worked has a clear answer; it isn't decided by whoever talks loudest.
Ownership is real
The team is looking for someone who treats the product as their own business, not someone in an execution-only role.
How to apply
Apply now
The online form takes about 2 minutes, and a Step1ne consultant will contact you as soon as possible.
Interview now with A-Cai, our AI interviewer Or ask us directly on LINE →