I'm planning to hire one or two research interns at MiroMind AI, in Singapore, on-site. Below is what I'm looking for, and why I'm writing the listing as an essay rather than a job description.
1. The shape of the work
In one sentence: working with me on agent systems. MiroMind's project space is broad — foundation models, agentic reasoning models, agent frameworks — so I won't pin you to a specific topic before you arrive. My preference is to let you spend a week or two reading what's currently in flight, then pick the project where your instincts feel sharpest, and go deep on that for a stretch.
The pace of the work will mirror how I work today: priority on shipping things that run, that get used, and that survive in production — not on producing papers. If a paper happens, it should happen as a natural consequence of work that reached enough depth. If the work hasn't reached that depth, forcing a paper out of it is wasted time for the intern and for me.
2. Who I'm hoping to hire
Hard requirements:
- Currently enrolled student. PhD, master's, or undergrad — I have no preference about level. From my own undergrad-then-PhD experience, undergrads can in some specific dimensions be braver than PhD students.
- CS-related major. This isn't a credential filter — the work itself requires it.
- Strong Python.
- Fluent vibe coding. That is, comfortable using Claude Code / Cursor / similar agents to write code — handing whole tasks to AI rather than typing each line yourself. This is not optional or a plus — it's a baseline tool for this work in 2026.
Softer, and more important:
- Self-driven. I won't supervise every step. Once you have a direction, you should be able to break the problem down, design and run experiments, produce intermediate results, and come tell me about them — without me having to ask.
- Genuine technical interest. Not someone treating the internship as a checkbox on a CV — someone who would still want to understand this stuff even if there were no credit and no paper at the end.
- A high bar. This is not about "long hours" or burning yourself out. It's about not accepting "the version that just barely works" as the endpoint. You should have a real internal sense of how good the thing in your hands actually is.
3. A few things I want to be explicit about
Don't come here to publish a paper. This is the most non-standard thing about how I hire. Many academic and industry mentors will tell you, the day you arrive, which conference your work is targeting. I won't. I'll tell you the level of depth this work needs to reach for someone to actually use it. If a paper falls out as a natural byproduct, ship it. If it doesn't, you've still produced something real — that's worth more than a forced submission.
If your goal is a paper for your CV, you'll probably be disappointed.
Doing matters more than presenting. Your first week here should be: get the dev environment running, read the existing code, ship a minimal demo. Not write a polished research plan. How you handle the raw problem tells me much more about fit than how you write a proposal — and tells you much more about whether this is the work you actually want to do.
4. Practical details
- Location: Singapore, on-site. I don't take remote interns.
- Timing: Rolling, no deadline. 3-6 months minimum is the suggestion; longer is better. Start whenever.
- Compensation: Paid. Numbers depend on the situation, but in the competitive range for AI internships in Singapore.
- Slots: 1 to 2.
- How to apply: Email me at bwang28c@gmail.com with your CV and links to your GitHub / open-source work. If there's a direction at MiroMind you specifically want to work on, or something you've been building lately, write a paragraph telling me about it — that paragraph often tells me more about fit than the CV does.
I reply to these myself. If you don't hear back in a few days, send a follow-up — it's not deliberate, just probably missed.