Software Engineer, AI Agents

MIAO WORLDS PTE. LTD.; Singapore; On-site; SGD 60,000 - 90,000; Full time

Job summary

MiAO AI, a Singapore-headquartered AI startup, seeks a software engineer to build concrete agent features and system modules on top of a custom agent architecture. You will translate requirements into production-ready code, adding new agent skills, tool integrations, verification workflows, and backend services.

You will join a small, elite core team, pairing with the Founding Engineer to rapidly grow your agent system design capabilities and production-grade experience.

Qualifications

• 1–3 years of software engineering experience with shipped products.
• Hands-on experience building agent systems or LLM-powered applications.
• Able to write code across the web tech stack (TypeScript, Python, Bash).
• Deterministic validation mindset before moving forward.

Responsibilities

• Build new skills, tool integrations, and product features on the existing agent framework.
• Deliver end-to-end features — frontend, API layer, and agent backend.
• Build and maintain verification pipelines for agent reliability.
• Debug agent failures across the full stack with production-quality code.
• Implement and maintain development infrastructure: testing, CI/CD, monitoring.

Skills

LLM-powered apps

Agent systems

TypeScript

Python

Bash

Tools

CI/CD

Monitoring

Job description

About MiAO AI

MiAO AI was founded in 2025 and is headquartered in Singapore, having raised over SGD 95 million from globally renowned investors. Our founding team previously built large-scale live-service gaming products with over 25 million DAU and 100 million MAU.

Our mission: Let AI discover you.

The next leap in AI isn't about getting smarter - it's about becoming yours. Seventy years of AI research keep proving the same thing: general methods win over clever designs. Models are getting stronger. Protocols are converging. Infrastructure is commoditizing. But one layer remains scarce - the layer that makes AI truly yours.

We're building personal AI agents - an AI that knows you better on day 30 than day 1, not because it got stronger, but because you shaped it.

Our workflow is 100% AI-driven - engineering, product, operations, no exceptions. No middle management. Radically flat. Every team member gets unlimited access to frontier AI models. We spend on efficiency, not process.

Who we're looking for: Believe in the Bitter Lesson - let AI discover, not discover for it.

About This Role

You build and deliver concrete agent features and system modules on top of the architecture designed by the Founding Engineer. You turn specific requirements into production-ready code - new agent skills, tool integrations, verification workflows, and backend services. This role emphasizes high-quality execution and rapid iteration.

You'll join a small, elite core team, pairing with the Founding Engineer to rapidly grow your agent system design skills. This is a role with a clear growth trajectory - you'll accumulate deep experience in production-grade agent systems.

What You'll Do

• Build new skills, tool integrations, and product features on the existing agent framework
• Independently deliver product features end-to-end - frontend, API layer, and agent backend as a complete chain
• Build and maintain verification pipelines to ensure agent reliability
• Debug agent failures across the full stack, writing production-quality code
• Implement and maintain development infrastructure: testing, CI/CD, monitoring, and observability tools

You May Be a Good Fit If You...

• Your daily workflow is 100% AI-driven - AI isn't an assistive tool, it's your default way of working
• 1-3 years of software engineering experience (or equivalent demonstrated ability), with shipped products - real software used by real users, not prototypes
• Hands-on experience building agent systems or LLM-powered applications - not experiments, but products serving real users
• Able to write code independently across the web tech stack (e.g., TypeScript, Python, Bash)
• Naturally verify before moving forward - your instinct is to run deterministic checks rather than trust the output