What we look for in Scholars

We say on the programme page that prior research experience is not required, and people reasonably suspect that is a polite formality. It is not. Here is the longer version.

We are selecting for a trajectory, not a résumé

A Scholar is with us part-time for a fixed period with a mentor. In that window, what matters is how quickly someone goes from not knowing something to building with it, because that is the whole shape of the work.

A candidate who taught themselves a domain over six months and produced something small and working tells us more than a candidate with a longer list of things they have been adjacent to. The first is evidence about how they learn. The second is evidence about where they have been.

Show us something you finished

The most useful thing in an application is a thing that exists. Not a large thing. A small tool, a script that solved a real annoyance, a bot that did something useful for a group you belong to.

Finished matters more than impressive. An enormous amount of the difficulty in agent work is in the last twenty percent: handling the case that breaks it, deciding what to do when the input is wrong, making it survive contact with someone who is not you. A candidate who has been through that once, even on something trivial, has learned something we cannot teach quickly.

We read for how you describe a failure

We usually ask about something that did not work. The answers separate sharply.

Some describe an external obstacle: the data was bad, the library was broken, the requirements changed. Sometimes true and rarely the whole story.

Others describe what they assumed, when they noticed the assumption was wrong, and what they did next. That answer is the one we are looking for, because operating agents is substantially the practice of noticing that a thing which appeared to work does not.

Curiosity about the boring layer

A candidate who is excited about model capability is normal. A candidate who is interested in what happens after the demo is rarer and fits what we do.

Who owns this agent. What did it cost. How do we know it still works. Can we prove what it did. Those questions are the substance of the Yaju platform, and someone who finds them interesting rather than tedious will enjoy the programme considerably more.

What we are not filtering on

Publications. Institution. Whether you studied computer science. Whether English is your first language. Whether your experience is professional or self-directed.

We are also not filtering on availability beyond what the programme needs. It is part-time and remote on purpose, so that people with other commitments can take part.

How the matching works

We consider background, skills and interests against the work available and make the best match we can. It is a judgement, made by people, and we try to make it a good one rather than a fast one.

If you are not selected, reapplying is welcome. We read it as a signal of continued interest rather than as a repeat.

Next

The Scholars Program page covers status, deadlines and benefits. Applications open in September each year. The Open Development Community is free to join in the meantime and is where a good deal of the work happens in the open.