Petri Dish Games · simulation, data and intelligence studio
Simulation, data and validation for healthcare systems
01
Capabilities
02
Where it applies
Each job starts as a root-cause investigation, tested against your data.
Capacity and flow
- Capacity studies: will a system keep up at peak hour, before you build or expand.
- Turnaround time: where a sample or result loses its minutes, from order to result.
- Fleet and staffing models: couriers, robots or remote monitoring teams.
Data quality and validation
- Data-quality checks that catch bad recordings and results early.
- Sensor and device accuracy against reference instruments, and when to recalibrate.
- Handling validation: do samples arrive intact, every route and every step.
- Environment monitoring: air quality, ventilation and pressure in patient areas.
Faults and security
- Fault analysis: where failures come from, traced to a root cause.
- Security review of control systems and device firmware.
- Chain of custody that never drops a record.
03
How a contract runs
04
Privacy and security
Built to your regulations
- The rules first: which laws apply to your data, devices and designs.
- Systems built to them, with the evidence your security team asks for.
- Secure devices: signed firmware, and testing that never touches a live hospital network.
How your data is handled
- An NDA before any data moves, and a business associate agreement wherever patient data is involved.
- De-identified data for development. Real data stays with you.
- Every AI tool named, and none trained on your data.
- Your data stays yours: returned or destroyed at the end.
05
Systems you own
- Runs on your own network and storage. Nothing depends on my servers.
- You own the code, the models and the data. No licence to renew.
- Independent of any hardware vendor, so the numbers answer your question, not a sale.
- Fits your existing stack, and your team can run it without me.
06
Proof from shipped work
My own ecosystem simulation is the testbed. The value is the automation around it: 17,004 changes in just over a year, each one checked automatically before it lands.
- Every change checked: 1,266 automated checks, and 113 recorded proofs that each check catches the fault it was written for.
- 20× faster, nothing changed: a simulation step cut from 23.7 ms to 1.2 ms, with 0 of 2,073,600 pixels different afterwards.
- No lost records: 50 programs writing at once, killed 20 times mid-write, nothing lost or half-written.
- Runs that replay exactly: the same inputs give bit-identical results, enforced by automated tests.
- Measured improvement: a search model tuned to put the right answer in its top 10 for 23% more questions it had never seen.
- Cost measured, not guessed: 371 test runs on 52 real past faults showed a cheaper AI model does the same jobs at about a quarter of the cost.
- Unattended operations: 5,142 automated jobs triaging reports, watching for crashes and benchmarking performance.
07
Background
- B.S. cellular and molecular neuroscience, Colorado State University, 2024.
- Electrophysiology research in the Reist Lab: Synaptotagmin mutations in Drosophila.
- Clinical EEG technician before founding the studio.
- Petri Dish Games: one neuroscientist directing a pipeline of AI agents. I design and verify; agents write the code.
08
Request a quote
Tell me about the problem. The quote comes back by email.
Or email max.pliskin@petridish.games