DeepBio Scientific builds AI systems for drug development. Helixir, our AI co-scientist, is the first of them: it generates novel hypotheses and de novo designs, drives them through a scientific harness of ML models and scientific tools, and returns evidence you can audit. We take a programme to the point where it is worth putting in a wet lab.
That is the whole reason the company exists. Most of what a programme needs to know already exists, scattered across literature, assays, structures and trials that no one team can hold at once. We build systems that reason over all of it, propose what nobody has tried, and hand back the evidence for and against it, so that fewer good ideas die unexamined and fewer bad ones reach a patient.
It is a long problem and we are not precious about who solves it. What we find that can be checked from outside, we publish. The measure of a year here is whether a disease is closer to having an answer than it was, not whether we shipped more than the company next door.
A scientific harness runs ML models and scientific tools together: generation, red-teaming, review and ranking, over a knowledge graph of ninety-plus sources. It is the spine under Helixir and under everything we build after it, so the engineering problem is orchestration that stays auditable while it runs for hours.
Representation learning, molecular design, evaluation and the biology underneath them. Some of it is published, more of it is in progress, and the direction is still being set. You would be choosing problems here, not picking them off a list.
Helixir, our AI co-scientist, is the first product built on the harness: chat, canvas, structured runs, a data hub and a living knowledge graph. More will follow across the drug-development stack. Joining now means shaping the second and third, not just maintaining the first.
These are the terms we hold each other to, and the ones we would hold you to.
A benchmark is not the finish line. The work is finished when someone is treated with something that did not exist before.
We publish what can be checked from outside, including the evaluations that make our own results harder to overstate. Knowledge that stays private helps nobody get better.
The diseases with easy answers already have teams. We would rather spend the decade on the ones that have resisted, because that is where the need is.
A disease-free world is a long horizon and no excuse for a slow week. People are waiting, which is a reason to move quickly and a reason never to cut a corner that matters.
Whoever brought it, whatever their title, however long they have been here. Ideas are argued with on their merits and nobody defends one because it is theirs.
Scientists stake programmes, budgets and careers on what we hand them. That is a responsibility before it is a market.
We are not posting roles right now, but we are always reading. If you work on orchestration, evaluation, or molecular representation, introduce yourself. Introduce yourself →
A CV or profile link is welcome. One honest paragraph tells us more than a cover letter, so keep it short.
DeepBio Scientific is an equal opportunity employer. We consider every application on its merits, without regard to race, colour, religion, sex, sexual orientation, gender identity, national origin, age, disability or veteran status. If you need an adjustment at any point in the process, tell us and we will arrange it.