The deal
The design-partner exchange
The exchange runs six months in four phases, with a checkpoint at the end of each. We are taking three partners.
The exchange
Four phases
| Phase | Weeks | What happens | Checkpoint |
|---|---|---|---|
| Scope | 1–4 | Your problem and success criteria, written down. | Both sides sign. |
| Build and baseline | 5–10 | Your baseline mesh and material properties, running on our simulator, expanded and tested byend of 2026. | Baseline reproduces a result you already know. |
| Run | 11–22 | Your counterpart runs your study on your hardware or cloud GPUs. | Stable under mesh and parameter change. |
| Report | 23–26 | Our numbers against your reference; hardware and wall-clock named. | Nothing left to clear: the report is yours to keep and publish. |
Cost
Same for every track
- License: 32 concurrent instances. An instance is one GPU card or one CPU chip.
- Compute: your own hardware, GPU or CPU, either on-prem or your leased cloud.
- Data: stays with you.
- Results: co-published, your sign-off.
- Design IP: yours.
Free for academia. $5k for startups and industry, covering up to 32 on-prem concurrent instances.
Market-segment exclusivity, from $2M
If the exchange ends with you ready to put the simulator into production, this is the next thing to weigh. The fee buys your licenses up front.
Every dollar is credited against on-prem instance licenses, which you draw down over several years, and for the exclusive year no other company in your named segment can license the simulator.
The terms
One named segment, one year from close. Bids set the fee, credited in full against on-prem licenses, priced per instance per year.
Segments: pharma, chemicals, materials and energy, semiconductors, quantum, fusion. Academic licenses are never exclusive.
Founder-led
Who you would be working with
The founder answers the email and writes the solver. Design partners decide what gets built next.
FOUNDER
Jorge Campos, Ph.D.
Ph.D., Electrical Engineering, UC Davis. Circuit-verification CAD, quantum device physics.
Nine years drafting patents.
Over ten years writing GPU code.
COMPANY
Bayris · numerical since 2016
- 2016–2022 AptML, a general-purpose GPU numerical framework, one source for GPU or CPU.
- 2023–2026 the multi-scale simulator, written on AptML.
- 2024–2026 a graph transformer for molecular properties and drug–target binding, in training.
ON FILE
3 patent applications
Filed, not yet granted.
- Multi-scale engine.
- Quantum specialization.
- Neural-network optimizations.
Apply
Apply to be a design partner
Tell us what you want out of the six months. The more specific you are, the faster we can tell you whether we can do it.