The design improves while we sleep.
We turn PET plastic into ethanol. This page is about how a small team does it quickly: two AI loops that run without being asked, one reviewing the reactor prototype every day and one refining the heat-recovery design for over a year, with more than a million simulation runs behind them.
AI loops improving the design around the clock
Monte Carlo simulations against the current reactor design
simulation runs before we touched the bench
Two AI engineers on the prototype. Zero new spend.
One AI designs and simulates the reactor. A second one tries to break every design. A person approves only the changes that are clearly better. The whole loop runs on subscriptions the company already pays for: no new hires, no new tooling, no idle nights.
The same design that goes to the print shop is the one the machines have already argued over.
- Daily
Full Monte Carlo simulations run against the current design and write results to a shared drive.
- Weekly
A scheduled AI session reads the drive, updates the CAD model and flags what the numbers say is wrong.
- Then
A person makes the call. The design advances only when a change is clearly better.
A model that has been improving itself for over a year.
Heat is the largest controllable cost in any conversion process. To get it right, Michael built a heat-recovery model that does not wait for a person to run it. It pulls from an external data source, evaluates a thermal-integration design for the process, and folds what it learned back into its own inputs for the next pass.
It has been running in the background since mid-2025 with no manual restarts. When the first plant is built, its thermal layout will already have more than a year of iteration behind it.
The data source and the model internals are confidential. The goal is not: a low-temperature process designed to give back most of the energy it uses.
- Pull
New data from an external source
- Evaluate
Thermal-integration design against the current process model
- Record
What improved, what did not, and why
- Feed back
Results become the inputs for the next cycle
One fuel operator, three fields of science.
The idea did not come out of a laboratory. It came from a fuel-logistics operator who knew what a gallon of ethanol was worth, how it moved through terminals and blenders, and how much post-consumer plastic gets buried every year.
Working through the published literature with AI as a study partner, Michael joined three bodies of work: engineered enzymes that break PET down at low temperature, immobilization methods that keep an enzyme working through repeated use, and continuous-flow reactor design that keeps plastic, catalyst and liquid in contact. None of the three was new on its own. The invention is how they are combined.
Every claim was tested against what the published science supports, which is why the provisional patent covers the immobilization system, the reactor and the integrated process rather than the enzyme.
- Jun 2025Concept formed and simulation work begins the same month.
- Jul 2025U.S. provisional patent filed. The heat-recovery loop starts running.
- Nov 2025PlastiBioFuel LLC formed in Texas. SDVOSB verified.
- Jan 2026Sponsored research agreement signed with a Texas research university.
- 2026Real bottle PET broken down on the first attempt, confirmed by HPLC.
Over a million runs, from flow rate to immobilization strategy.
Before committing to hardware, Michael built simulation models that reproduce the numerical methods used in leading engineering software: computational fluid dynamics for flow and residence time, reaction-kinetic models for enzyme behavior, and process models for mass and energy balance. Then he ran them more than a million times.
Each family of runs narrowed the design space for the next, so by the time work moved to the bench, the experiments were confirming a design rather than searching for one. That is why the first attempt on real bottle plastic worked.
Flow
Channel geometry, velocity and dead-zone elimination modeled across the operating envelope before fabrication.
Kinetics
Rate, turnover and substrate depletion modeled against the crystalline PET a plant will actually see, not idealized film.
Immobilization
Loading, retention and reuse strategies compared head to head to extend enzyme life and lower cost per ton.
Thermal
Temperature profiles and heat integration modeled to hold the process at low temperature with the smallest energy input.
Pretreatment
Size reduction, washing and contaminant handling for the mixed, dyed and dirty grades recyclers reject.
Separation
Ethanol recovery and water recycle modeled as one system so the mass balance closes.
The lab decides.
- Measurement over simulation. Simulation proposes; chromatography, activity assays and mass balance decide. No claim leaves the company without a measurement behind it.
- Novelty is scoped precisely. PlastiBioFuel did not invent PET-degrading enzymes. Our work is the immobilization system, the reactor and the integrated process.
- A person owns every output. Proposals, contracts and technical documents are reviewed and signed by the founder.
- Proprietary detail stays proprietary. Materials, operating parameters, data sources and model internals are not published here. This page describes the approach, not the recipe.
Questions.
What does PlastiBioFuel make?
Fuel-grade ethanol made from post-consumer PET plastic, the plastic in water and soda bottles, using an immobilized enzyme that works in warm water.
Does PlastiBioFuel use AI to design its process?
Yes. The process was conceived, simulated and refined with AI as a working partner, including more than one million simulation runs before the first bench experiments. Laboratory measurement decides what is carried forward.
How many simulations has PlastiBioFuel run?
More than one million runs across flow, kinetics, immobilization strategy, thermal design and separation, using numerical methods that match leading engineering simulation software.
What is the prototype review loop?
An automated design-review cycle for the reactor prototype, built on tools the company already subscribes to. One AI designs and simulates, a second stress-tests every design, and a person approves only changes that are clearly better. Monte Carlo simulations run daily against the current design, and a scheduled AI session updates the CAD model each week.
What is the heat-recovery loop?
A self-building thermal-integration model that has run continuously for more than a year. It draws on an external data source, feeds each result back into its own inputs, and improves the heat-recovery design for the process without manual restarts.
Is the technology protected?
A U.S. provisional patent application (63/848,456) covers the enzyme immobilization system, the reactor and the integrated process. Specific materials and operating parameters are confidential.
Is PlastiBioFuel a veteran-owned business?
Yes. PlastiBioFuel LLC is a Service-Disabled Veteran-Owned Small Business (SDVOSB), verified through SBA VetCert, founded and led by a 12-year U.S. Army veteran and based in Fort Worth, Texas.
How can I work with PlastiBioFuel?
Email michael@plastibiofuel.com, book a 20-minute call, or use the contact page. The company works with research partners, fuel buyers, waste-stream suppliers, federal program officers and investors.
See the process up close.
Collaborators, program officers, fuel buyers and investors are welcome.