Join us

We are hiring research-automation researchers.

We have repeatedly had an experience roughly like: we notice a way that AI-driven research feels like it’s failing to be excellent, think about what kind of scaffolding might help to address that, build it — and it often pretty much works. There are a bunch of low-hanging fruit for this kind of manual exploration (corresponding to a bunch of ways in which AI-driven research isn’t there yet), and we are capacity-bottlenecked on exploring them all.

Working out how to make research automation go well — research-automation research — looks like an increasingly important field over the next few years, and we want to accelerate it and set it up for success. We are focused on macrostrategy: thinking about the position the world is in, and where it is going.

Role
Macrostrategy research-automation researcher
Seniority
Open — we’ll work with people to get them set off in good directions
Salary
$200,000 – $300,000+ (USD)
Equity
0.5% – 5%, on top of salary. 5% is what the founders have
Location
Anywhere in the world. We may have occasional team retreats
Visas
We cannot sponsor visas at present
Apply to
jobs@differential.tech

The work

Fundamentally, this is a high-autonomy research role — we are looking for partners who can help us to make progress and shape the field for the better. But we are open to people at varied levels of seniority stepping into this, and we’ll work with people to get them set off in good directions.

Our work includes:

  • Building scaffolds that address one part of the research pipeline, ideally modular so that they may be invoked by other scaffolds.
  • Refining and testing variations of existing scaffolds.
  • Building and refining metrics and evaluations for assessing performance at various research tasks.
  • Building knowledge infrastructure, towards a future where we have something like a digital analogue of peer review, or other institutions.
  • Developing an emerging science and set of best practices around how to automate research.
  • Packaging useful capabilities into deployed systems which may be helpful for users.

In practice, the work will be a mix of:

  • Reading outputs, or intermediate outputs, from scaffolded systems, and assessing what is high-quality versus what is weak.
  • Prompt-engineering, or designing new scaffolds to try to reliably elicit certain types of behaviour.
  • Building systems to automate parts of this.
  • Meta-level planning and communicating with others on the team.

Macrostrategy research experience would be a plus. Software engineering experience would be a plus. Neither is necessary.

This role tracks what the emerging field needs. It sits somewhere between traditional research — ideally in the service of dogfooding the systems we are building, you will spend 20–60% of your time engaged in object-level research — software engineering, and management. In many ways, building scaffolds for LLM instances amounts to trying to develop a practice of research management for automated research.

We are a small team and expect versatility to be more important than specialization at this stage, although we are not against hiring specialists if there are strong candidates. For the right person, we’d be excited to have them act as org leadership as we continue the work and, we anticipate, scale up.

Culture

Things we deeply believe in:

  • Assembling a team of people who earnestly want to do good things — who will be happy to argue with us when things seem wrong, and happy to be responsive to arguments.
  • Respecting and trusting our staff. In practice, that means we’d rather support you in making good choices about what to work on than tell you what to do. We think we’re further along this than most places, and we intend to stay that way as we grow.

Right now, we have vision for things we want done — but our greater excitement is to find and empower people who want to see the same things we do.

You can do this work from anywhere in the world. We may have occasional team retreats. We cannot at present sponsor visas.

Compensation

We aim to offer competitive compensation. Salary is typically 200–300k. We’re prepared to go above this range for exceptional candidates.

As well as salary, anyone joining at this stage should expect to get between 0.5% and 5% equity — 5% is what the founders have.

We wish to stress that we are not primarily aiming to make money, and may never seek to commercialize — but we think it’s appropriate that if we do ever commercialize, staff are appropriately compensated for this.

Applications

If you’re potentially interested, please reach out to jobs@differential.tech. If you don’t already know one of Owen and Lawrence, please provide the name of some references whom we might speak to about you.