Apex is one of a small number of labs in the world training a foundation model from scratch. We sit alongside the chat-trained frontier labs as the third axis — focused on the physical world rather than language. Aether is the foundation model we train end-to-end on simulation trajectories, instrument data and design history, and ship into the disciplines that have to face reality.
Aether is the foundation model at the centre of the company. We train it from scratch on simulation, instrument and design data — not on scraped text. The same backbone serves every discipline we touch.
The short version of an answer that takes longer in person. Long enough to be honest, short enough not to be a manifesto.
The next decade's hardest problems — climate, drug discovery, energy, semiconductors — share a substrate: the physical world runs on physics, chemistry, biology and silicon. A model that understands one understands the rest. Building five sibling models is a strategy mistake.
The seven-figure software stacks at incumbent vendors are mature, hard to dislodge and overdue for replacement. We don't compete with the incumbent CAE, EDA and computational-chemistry suites by being marginally cheaper. We compete by collapsing their entire shape into one model with one contract.
Foundation models for physical work have a five-year head-start over their incumbents. We are spending those years training on data nobody else has access to, while the incumbent stacks ossify on per-seat licences and Tcl scripts.
These are the principles we hire against and ship against. They are short on purpose — long principles are decorative.
We are training one AI on simulation, instruments and design history. Not wrapping someone else's chat API and calling it an agent. The model is the company; everything else is product surface.
Every claim ships with a benchmark you can rerun. If we can't reproduce it, we don't publish it. If the comparison isn't fair, we say so on the same page.
We prefer projects where the model has to face physical reality — wet labs, fabs, test rigs, taped-out silicon. Reality is the regulariser. There is no substitute.
Biosecurity, export-control and IP boundaries belong inside the model, not in a wrapper that someone can disable. We red-team every release against a versioned corpus before it ships.
Production systems beat demos. We invest in the unglamorous middle — queues, retries, traces, rollback, on-call runbooks. The demo is a poster; the boring infrastructure is the product.
Every failed run, every bad mesh, every off-target hit is institutional memory. We build platforms that remember — for the customer and for ourselves.
We're young, on purpose. Apex is built to make decisions on a five-year horizon, with the urgency of a company that is.
Apex founded. First simulation-pretraining checkpoint runs end-to-end on a single rack.
First customer pilots in aerospace CAE and small-molecule discovery. Aether preview released to academic partners.
Aether for drug discovery and semiconductor disciplines added to the platform. unlimited workloads across engineering reach production.
Aether 1.0 generally available. Three sizes — 7B dense, 40B dense, 280B sparse-MoE — with 1M-token context and streaming rollouts.
Sovereign deployments at national-lab consortium. First autonomous-lab campaign in ALCOA+-compliant production.
A small, deep bench. Researchers who have shipped models at scale, applied engineers who have shipped pilots, and an operations team that takes the boring parts seriously.
Founder and CEO of Apex. Bet the company on a single thesis — the next decade of AI for the physical world runs on simulation, not language, and it deserves a foundation model trained from scratch. Sets research direction, owns hiring, signs every customer pilot, and is on the hook for every release that goes out.
Co-founder and Chief Scientist. Twenty years of numerical methods across fluid, structural and quantum systems. Previously held appointments at two research universities; forty-plus peer-reviewed papers. Owns the technical direction of the Aether model family — the corpus, the architecture, the cross-domain pretraining programme that makes one model serve every discipline.
Built and operated distributed-systems infrastructure at hyperscale before joining Apex. Owns the agent runtime, the deployment substrate and the production security posture. The on-call rotation runs through her; the customer-impact bar on production incidents is hers to hold.
Runs the research organisation and the publication programme. Previously led a physics-informed-ML group; published widely on simulation-augmented learning and long-rollout stability. Owns the research agenda, the eval suite and the public-benchmark commitments — including the ones that would have been easier not to publish.
Runs the forward-deployed engineering team — the engineers who pair with customer teams and land the pilots. Background in industrial CFD and aerospace systems; has personally retired three seven-figure CAE renewals at customer sites. Believes pilots are won engineer-to-engineer, not in slideware.
Owns the refusal corpus, the capability-gating architecture and the external red-team programme. Previously led safety engineering at a major AI lab. Acknowledged contributor to the open biosecurity-evaluation literature; chairs the external safety advisory board's working sessions.
Owns the model-serving substrate, the eval-gated deploy pipeline and the observability stack. Previously built large-scale training infrastructure across two foundation-model labs. On-call lead for production Aether — the runtime is hers to make boring.
Runs the engineering discipline — CFD, FEA, electromagnetics, multiphysics, additive, fatigue. Twenty years across aerospace and automotive CAE. Has personally retired three seven-figure CAE renewals at customer sites; the engineering accuracy bar inside Apex is his.
Runs the Aether for drug discovery. Computational chemistry and structure-based design background; previously led discovery platforms at two clinical-stage biotechs. Owns the loop from in-silico hit to wet-lab confirmation — and the handoff between Aether for drug discovery and Aether for autonomous labs.
Runs the Apex semiconductor discipline — RTL through signoff on one database. Two decades of EDA across major foundry process nodes; holds half a dozen patents in physical-implementation tooling. Believes the next chip belongs to whoever closes signoff fastest with the smallest team.
Owns the studio surfaces — the canvas, the review threads, the agent traces engineers actually look at. Previously shaped design at two productivity-software companies known for unusually careful interfaces. The thing engineers see every day is hers.
Runs commercial. Background in enterprise software where renewals are seven-figure and procurement cycles are long. Believes pilots are the only honest sales motion — and the comparison memo, co-authored with the customer, is the real contract.
The bench of senior researchers driving Aether. Each researcher names the area they lead, recent papers, and one open problem they would like to hear from you about.
We hire small numbers of unusually senior people, give them unusually large scope, and ship unusually large results. Remote-friendly, in-person-preferred, compensation honest.
We publish bands in every job description and we do not haggle the offer down. Equity and benefits below; cash bands on the role page itself.
Top-of-market base, honest range published in every job description. We do not negotiate against ourselves; we make our first offer the best offer we are prepared to make.
Refresh grants every two years, accelerated vesting on involuntary termination, transparent strike-price disclosure, ten-year post-termination exercise windows.
Four weeks paid vacation, no questions. Sabbatical at year five. Real, not 'unlimited and never taken.'
Premiums covered for the employee and dependents. Mental-health budget per employee per year. Coverage carries into a 12-month bridge if you leave.
For interviews, brand assets, the company boilerplate or to verify a number before you publish it — please email press@apexworldlabs.com. We aim to respond within one business day.
Apex is a frontier AI lab — one of a small number of labs in the world training a foundation model from scratch. Aether is the simulation-native foundation model at the centre of the company: trained on solver trajectories, instrument data and design history, it runs forward in time across fluid, structural, electromagnetic, quantum, biological and digital systems under one set of weights. While the chat-trained frontier labs focus on language, Apex focuses on the physical world — engineering, semiconductors, drug discovery, autonomous labs, materials, fusion and software-engineering work. Founded in 2024 and headquartered in San Francisco.
Marks, wordmarks, palette, typography specimens, and the usage rules — open on /brand. Print-grade variants on request.
Open roles, custom roles, deferred roles. Send a note even if you don't see a perfect fit.