AI Adoption Pod Engineer Intern

Carousell Group; Singapore; On-site; SGD 33,000 - 50,000; Full time

Job summary

Carousell Group is building a small pod of engineers to help internal teams figure out the right tools, workflows, and agents to multiply their impact. This isn't about basic build support — most teams can already put together a simple AI workflow on their own.

The pod exists for the harder calls: what should run on Claude versus another tool, how to weigh cost and latency, how to architect the connection between a front-end tool and the underlying infrastructure so it holds up under real use,

Qualifications

• Strong fundamentals in software engineering — writes correct, working code independently rather than completing a guided assignment
• Strong working knowledge of GenAI primitives — prompting, context engineering, MCP tool/function calling
• A track record of shipping something real end-to-end (personal project, academic project, or internship)
• Given an ambiguous, unscoped problem, can independently break it down and drive to a solution without a detailed spec

Responsibilities

• Sit with internal teams to understand what they're trying to solve — sharpens fuzzy problems through conversation and rapid prototyping
• Build and iterate on AI-powered skills, workflows, and agents for real, non-technical users
• Make the calls a non-technical builder can't: what to deploy and where, how to weigh cost against speed and latency
• Debug in production when something breaks for a real user
• Stay with a workflow past "it's built" — ensure it works and outcomes are measurable
• Feed patterns back to the pod: reuse across teams or tailor as needed

Skills

Software engineering fundamentals

GenAI primitives experience

End-to-end shipping

Problem decomposition

Education

CS/Engineering degree or student

Tools

Claude

LLM tooling

Job description

Carousell Group is building a small pod of engineers to help internal teams figure out the right tools, workflows, and agents to multiply their impact. This isn't about basic build support — most teams can already put together a simple AI workflow on their own.

The pod exists for the harder calls: what should run on Claude versus another tool, how to weigh cost and latency, how to architect the connection between a front-end tool and the underlying infrastructure so it holds up under real use,