Cemvita and Roebling Bring AI-Native Engineering to Industrial BioSolutions

Industrial biotechnology has no shortage of promising discoveries. The harder question is whether those discoveries can become industrial projects that work economically at scale.

Feedstock availability, fermentation performance, downstream processing, energy requirements, plant location and product demand all influence whether a biological process can compete in the real world. Yet some of those questions may not be fully explored until years into development, when changing direction becomes expensive.

Cemvita and Roebling want to bring those decisions forward.

The companies announced a partnership to combine Cemvita’s industrial biotechnology expertise and proprietary biological and operational data with Roebling’s AI-native process engineering and techno-economic modeling. The objective is to evaluate technical and commercial pathways earlier, helping identify which BioSolutions are worth advancing and what must change to make them viable.

“The bottleneck is not producing one more exciting result in the laboratory. It is turning that result into a deployable BioSolution,” said Moji Karimi, CEO and co-founder of Cemvita.

A deployable BioSolution, Karimi explained, has to work within a complete industrial system: using available feedstock, operating at the right scale and location, producing something customers want, and generating economics that can attract capital. Cemvita wants those deployment realities to influence biological development from the beginning rather than arrive as expensive surprises later.

Connecting biological performance to plant economics

Cemvita, founded in 2017, will contribute data and operating knowledge spanning feedstock composition, biological performance, fermentation conditions, productivity, yield, downstream recovery, utilities and scale-up behavior. Karimi said that includes what the company has learned moving from the laboratory through pilot operations and 75,000-liter scale.

In May, Cemvita completed a 75,000-liter fermentation campaign at Bio Base Europe Pilot Plant in Ghent, Belgium, producing FermOil™ from crude glycerin, a byproduct of biodiesel production. The company reported consistent results from 30 liters to 75,000 liters, a 2,500-fold increase in volume.

That information matters because a strain or process cannot be evaluated on biological performance alone. Yield improvements, for example, need to be understood alongside recovery requirements, plant configuration, energy consumption and the cost of operating the resulting facility.

Karimi also highlighted the value of learning from projects that have struggled. Feedstocks can change, downstream processing can prove harder than expected, operating costs can rise, and facilities may not adapt to new commercial conditions.

“The successes are data. The failures are data too, and often more valuable,” he said.

Today, said Edward Shenderovich, CEO and co-founder of Roebling, biology, engineering and finance often live in separate worlds: a biology team measures strain performance, an engineering firm builds a design months later, and an analyst builds a spreadsheet on top. Roebling’s approach puts them in one model, connecting biological performance to mass and energy balances, equipment sizing, capital expenditure (CapEx) and operating expenditure (OpEx).

“Biological performance goes straight into the mass and energy balances, the equipment sizing, the capital cost and the operating cost,” Shenderovich said. “The calculations are deterministic, not guesses from a language model.”

The practical result, he said, is speed with rigor. “A question that used to need a new study can now be answered in minutes. Teams can test many options instead of committing early to one.”

The model can also work in reverse: by identifying which biological improvements would have the greatest effect on a project’s economics, it can help direct research toward the changes that matter most commercially.

A common model for multiple deployment scenarios

For Cemvita, the initial work will establish a common baseline from existing design, engineering and economic information for its lead industrial BioSolution program. From that foundation, the companies intend to evaluate different plant capacities, feedstocks, locations, process configurations and economic assumptions.

Cemvita’s FermWorks™ plant provides a concrete example of the challenge. It converts carbon-rich waste streams and industrial byproducts into FermOil™, FermSource™ and FermNatural™. Cemvita has a portfolio of potential projects around the world, each with its own feedstocks, utility costs, scale and product requirements.

Traditionally, rebuilding the engineering-level analysis for every configuration could take enormous time and expense. Karimi described the new approach as a common baseline model, almost like a digital twin or world model of the plant, with each new opportunity treated as a scenario rather than a fresh start.

That could help Cemvita compare opportunities on a more consistent basis, identify the assumptions driving value, and determine what requires further validation before committing development resources and capital.

The same approach is intended to support earlier-stage programs as well as more advanced projects. When a customer project moves into front-end engineering, Roebling may also build a dedicated model to support that work.

From faster analysis to better investment decisions

Cemvita expects the partnership to shorten parts of the analysis informing front-end loading (FEL) decisions from months to weeks. The claim is deliberately scoped to portions of early analysis, rather than the entirety of engineering, development or deployment.

The commercial value is not simply speed. Industrial projects must contend with changing feedstock prices, product demand, energy costs, policies, credits and market premiums. A design that looks attractive under one set of assumptions may become less viable when those conditions change.

“The goal is not to predict one future perfectly,” Karimi said. “It is to build enough flexibility into the design that the project can succeed across several plausible futures.”

A faster way to examine those scenarios could help teams compare more pathways before committing significant capital, identify the variables that most affect viability, and eliminate weaker opportunities earlier.

The companies are not yet publishing a quantified reduction in project cost or risk. Karimi said establishing defensible performance benchmarks will be an objective of the initial work.

For Shenderovich, the larger test is whether the approach becomes repeatable. Over the next 12 months, Roebling wants to see Cemvita’s lead program running on its model and informing real deployment decisions, at least one customer project advancing into funded front-end engineering, and Cemvita using the platform as standard practice to screen new BioSolutions before spending heavily on them.

Industrial biotechnology’s scale-up challenge is not only a question of whether biology works. It is whether the entire process can work at a particular scale, in a particular place, under real commercial conditions. By connecting biological performance with engineering and economics earlier, Cemvita and Roebling are aiming to make that question easier to answer before the most expensive decisions are made.