Safe Superintelligence Jobs (NOW HIRING)

Physical Superintelligence is an exceptionally well-funded seed-stage startup with roots at Google, Meta, DeepMind, NVIDIA, Citadel, Harvard, MIT, Johns Hopkins, the Perimeter Institute, and the Institute for Advanced Study, building AI systems to discover new physics at scale.

Our mission is to discover and commercialize transformative physics breakthroughs at scale with artificial superintelligence, safely, verifiably, and for broad public benefit.

Our platform runs discovery campaigns against open problems in physics. Those campaigns produce results, and someone has to decide what happens to them.

That turns out to be harder than it sounds. A result you put in a paper may be a result you have given away. A result you put in a patent is one you cannot talk about for a year and a half. Plenty of results that look like breakthroughs have no patentable subject matter in them at all, and telling those apart quickly is most of the skill. We need someone who owns the portfolio of problems we point the platform at, sets the bar a result has to clear before anyone outside the building sees it, and works out — case by case — whether it should become a publication, a filing, a trade secret, or open source. The final call on the genuinely hard ones sits with a standing committee. The recommendation, the reasoning, the record, and the follow-through are yours.

You are not the physicist on these results and you are not the lawyer on these filings. You are the person who turns what the platform produces into assets that last, instead of a pile of interesting outputs.

Point the platform at the right problems. You will map the landscape of open physics problems our system can actually reach and decide where to spend campaign time. That means running the cycle of cheap broad probes to find fertile ground, then concentrating effort where the ground turns out to be fertile, then killing what is not converting. The best problems scientifically are usually not the best problems commercially, and you will be making that trade constantly, in front of people who care about it. The portfolio is a live allocation decision, not a wish list, and you own defending it.

Decide what happens to what comes out. Some results are obviously scientific and some are obviously commercial, but most are both or neither, and that is where the value and the mistakes are. The way through is a written rule set that handles the recurring patterns so that only genuinely novel cases go to the committee, and so that fewer cases are novel each quarter. The routing question is not whether something feels scientific — our best results will be both. It is whether there is eligible subject matter at all or only a law of nature with the value sitting in a downstream claim; whether a competitor could reverse‑engineer it from what we ship, since a trade secret protects nothing if they can and a patent just teaches them if they cannot; whether a contract we signed before the result existed has already decided the question; and whether we could realistically detect and pursue infringement. You will read our commercial agreements as closely as you read our physics. One rule has no exceptions: filings come before disclosure, because most of the world grants no grace period and an early preprint gives away international rights for good.

Get results validated, then get them published. You will build a standing bench of outside experts across the subfields our campaigns touch — people whose review actually means something — and route provisional results through them before anything goes further. That network needs infrastructure to be safe: NDAs, export screening, conflict screening, and a protocol that keeps a reviewer from contributing an inventive idea and clouding title. Then you take what survives to strong venues and manage it through submission, referees, and revision. Publication is how the field learns to take us seriously, which is worth a great deal to us. You are accountable for what the machine produces, not for being on the byline.

Build the IP pipeline. Right now there isn't one. You will stand up disclosure intake, triage, and a filing cadence, and direct outside counsel on strategy and priority while they draft and prosecute. You will keep the portfolio pointed at where value is actually captured rather than where papers are easiest, hold the boundary against our open‑source releases, and keep it in a shape an investor's technical diligence team can follow without you in the room. Part of this is unglamorous and load‑bearing: current USPTO guidance requires that a natural person significantly contributed to the conception of an invention, and the more of the work our platform does, the more that record matters. It has to be built into how campaigns run, because it cannot be reconstructed later. If nobody does it, we end up with papers and no patents.

Real technical depth in physics or an adjacent field. Enough to read a campaign output and judge whether it is real, novel, and significant, and enough to hold your own with our research team. A PhD is the most common way to get there and it is not the only one; a serious research career, or years of hands‑on technical work close to the physics, can do the same. What we cannot work with is someone who needs the result explained to them before they can classify it.

A technology transfer and IP management background. You have run the path from result to filing to publication yourself, with real stakes, and you know the places it goes wrong. Knowledge or technology transfer at a national lab, a research institution, or a university office; in‑house IP strategy at a deep tech company; managing a prosecution docket and the counsel attached to it. Advising on this from the outside is not the same thing.

Portfolio judgment. You have run a research program with a real prioritization mechanism and a real kill mechanism, under a budget and against a clock. You can tell us about something you shut down while it was still interesting.

A network, not a Rolodex. You have standing relationships with credible people across several physics subfields, you know who is worth listening to and who is merely loud, and those people will take your call.

You know when there is no patent to be had. You can look at a result and tell whether there is eligible subject matter in it or whether you are looking at a law of nature in an apparatus costume. You have had that argument with a scientist who was sure their discovery was patentable, and you were the one who was right.

The nerve to hold a line you cannot enforce by title. You will be under pressure from every direction to let something out early, and the pressure will be reasonable every time. Your leverage is the rule set and your credibility, not your seniority. You have held a line like that before without losing the relationship.

Clear eyes about AI‑generated science. You know where these systems produce insight and where they produce plausible nonsense, and you treat verification as the thing that separates the two. You do not need to build the platform. You do need to be unimpressed by it until it earns otherwise.

• Editorial or program committee experience at a major venue.

• Experience with export control compliance and sensitive technology research and development.

• Named inventor on issued patents.

• Familiarity with the commercial domains we work in: compute infrastructure on the ground and in orbit, energy, and sensing and instrumentation.

We are engineering‑led. Engineers own problems end to end, from spec to ship to on‑call. We write contracts before logic, test against real systems instead of mocks, and favor simple designs that ship over clever ones that do not. Our development process is AI‑native and we expect the discovery process to run on the same principles: instrumented, agent‑augmented, and measured.

We are a Public Benefit Corporation, and we run an academic grants program. The public benefit commitment is not decoration — it is a live input into the decisions this role handles, and you should expect to argue about it.

This role is based in Boston and is on‑site at our Cambridge hub. We offer competitive compensation including salary, benefits, and meaningful early‑stage equity. We evaluate judgment, what you have built, and the leverage you create.

This role involves access to controlled unclassified technical data, and some responsibilities are subject to U.S. export control regulations; we will discuss any resulting requirements transparently during the process. We are an equal opportunity employer and value diverse perspectives in building platforms for AI‑driven discovery.