Forward Deployed AI Engineer Intern (6 months)

PRISM+ is Singapore’s leading homegrown electronics brand, with a presence across key markets in Australia, Malaysia and the Philippines. The brand’s mission is to make premium technologies accessible to the masses—disrupting established industries with innovation and value that goes beyond the product. PRISM+ was founded in 2017 as a direct-to-consumer (D2C) display technology specialist, where it quickly grew to become the number one monitor and smart TV brand in Singapore. Best known and loved for its affordable premium offerings and award-winning customer experience, the brand has expanded its offerings to include a wide range of affordable premium home electronics and appliances.

Join us at PRISM+ and become part of a journey where innovation meets impact, and every day offers a chance to shape the future of consumer technology. Discover a career where your ideas matter, your growth is nurtured, and together, we create a brighter tomorrow.

About Orbit
Orbit is PRISM+'s forward deployed AI team. We're a small team embedded across PRISM+, helping each department choose the right tools, workflows and agents for how it works. Setting up a simple AI workflow is getting easier for every team. The harder part is the decisions around it, and that's where Orbit comes in:

• Which tool a problem needs: Claude, Codex, a plain script, or no AI at all.
• What it should cost to run, and how fast it has to be.
• How it connects to our data and systems so it keeps working after the first week.
• Who gets access, how it stays secure, and who looks after it once we hand it over.

The Role
As an intern, you'll join Orbit and work inside one business team at a time, next to the people who own the process. They're your customer. You'll find where AI can help them most, build it, and stay until the team is using it. You'll start in Operations/Service, Retail, Finance, Product Engineering or Marketing, and you can move to a different team for later projects.

How a project runs

1. Shadow. Watch the work being done and find out where the time goes and what goes wrong. The problem you're handed is rarely the real one.
2. Scope. Agree on the smallest version worth building, and how you'll measure "better".
3. Build. Prototype with code and AI, using tools like Claude, Codex, Slack, Notion and Google Workspace. You decide where it runs, what data it can touch and who has access.
4. Launch. Put it in front of users, train them and fix what breaks. If it fails for a user, it's yours to fix.
5. Hand over. Write a one-page guide, pass it to someone in the team, and report what it saved.

Our definition of Done
A project is done when

• the team uses it every week without being reminded
• the improvement shows up in numbers, like hours saved or errors avoided
• someone in the team can run it and handle small fixes without you

Your six months

• Weeks 1–2 - Foundations. Tools, guardrails, security and data handling. Start shadowing your first team.
• Weeks 3–10 - Project 1, in your first team. A narrow, self-contained problem, live and in use by week 10.
• Weeks 11–20 - Project 2, in the same team or a new one. A larger problem, and you manage the relationship with the team yourself.
• Weeks 21–24 - Project 3 in another team, or hand over. Take on one more problem, or harden what you've built and train the people who'll run it.
• Weeks 25–26 - Final review. Present what you shipped and what it saved.

How the process works
About three weeks, five steps. You'll always know which step you're on.

1. Apply (20 min) — a one-page CV and four short questions.
2. Screening call (25 min, video) — five fixed questions.
3. Take-home (4 hours, hard cap, 5 days to return) — a realistic messy problem from your target department. Use any tools including AI, tell us how you used them.
4. Panel (30 min) — walk us through your take-home, think through a fresh problem out loud, and meet your would-be department head in a short role-play.
5. Final (30 min) — terms, commitment, and a couple of questions.

Requirements

• Coding ability is a great to have - You write working code on your own, and you can read someone else's script and change it without breaking it.
• Proof you've shipped - At least one project (personal, academic, club or work) that someone other than you actually used.
• Business sense - You can see how a process affects cost, customers and time, and you go after the problem that matters most to the business.

• Hands-on AI building - You've built something that uses an AI model through prompts, context or tool calling, including at least one agent or multi-step workflow that uses tools. You should have hands-on experience with Claude, Codex or another LLM.
• Judgment about AI output - You check what a model gives you, can explain how it might be wrong, and know when a script or spreadsheet would do the job better.
• End-to-end understanding - You understand how a front end, an API and a database fit together, and can connect them yourself.
• Live debugging - You've tracked down and fixed a bug in something running live.
• Trade-off thinking - You can explain why one design is cheaper, faster or safer than another.
• Clear communication - You listen before suggesting a fix, and explain your choices to non-technical colleagues in plain English.
• Commitment - Six months, full-time, starting between November 2026 and January 2027.