A larger reactor does not automatically reproduce a smaller batch

On October 7, 2026, Samsung Biologics reported that Yeonjeong Choi of its New Modalities Group presented computational fluid dynamics (CFD) work on antibody–drug conjugate (ADC) scale-up at BIO JAPAN in Yokohama. The company said the modeling supported consistent ADC quality at commercial production scale.

But using a larger vessel may alter how material flows, where the payload enters, and how long it takes to reach a homogeneous state. Successfully producing a small batch is not by itself evidence that every relevant physical condition will transfer unchanged.

Samsung's release does not disclose the full experimental dataset or the regulatory status of any specific customer product. The more useful analytical question is how physical predictions, experimental validation and manufacturing quality evidence relate—without upgrading a company announcement into an independent GMP certification.

Why antibody–drug conjugation is also a mixing problem

An ADC uses an antibody to target cells and a linked drug payload to deliver a biological effect. Its conjugation process must control how antibody, linker–payload and reaction conditions produce the intended product attributes.

Samsung's bioconjugation service identifies control of the drug-to-antibody ratio (DAR) profile, conjugation parameters and analytical methods as development priorities. An average DAR alone does not establish that every molecule has the same distribution or that residual materials are within limits.

When the payload is added, local concentrations may differ near the feed point before the vessel becomes well mixed. CFD can estimate fluid velocities and material transport. It therefore helps answer not merely what came out of a batch but what happened in space and time while the chemistry progressed.

Three times in a separate study: 9.4, 32.2 and 17.6 seconds

A 2023 paper by researchers at Karlsruhe Institute of Technology and AstraZeneca combined conjugation experiments with CFD and kinetic modeling. Its reactor simulations covered three different vessels, up to approximately 50 L—not Samsung Biologics' commercial facility.

The predicted global mixing time was 9.4 seconds for the GST-1 glass stirred tank, 32.2 seconds for another glass vessel, GST-2, and 17.6 seconds for a single-use mixer, SUM. GST-2 and SUM had similar liquid volumes yet substantially different homogenization times.

The researchers linked those differences to impeller arrangements, stirring speed, and the axial/radial transport of fluid. These are specific modeled values for specific vessels, not measured performance claims about Samsung's 500 L infrastructure.

Different mixing speeds need not mean different final DAR

The same publication coupled CFD with a kinetic model for a site-directed DAR 2 conjugation. It found localized differences primarily during payload addition. Mixing lasted roughly 10–30 seconds across the modeled systems, whereas the relevant reaction times were approximately 300–900 seconds.

For this particular chemistry and studied parameter range, final DAR values remained largely unchanged. That is an important counterpoint to a simplistic 'slow mixing equals poor product' story: scale-up risk depends on the relative timescales of mixing and reaction, not reactor size alone.

The result cannot be generalized to every payload, linker chemistry, mixing geometry or commercial process. Other reaction kinetics or feed conditions may change the outcome. The study supports a way of asking and testing the question, not a blanket guarantee of ADC equivalence.

A flow simulation cannot replace a quality test or GMP validation

CFD predicts conditions inside a vessel. Product analytical methods must separately establish whether manufactured ADC meets its quality specifications, and GMP systems must demonstrate that processes and controls are appropriately qualified and operated.

Samsung's service catalog lists DAR, residual solvent, free drug, antigen-binding and cytotoxicity assays as distinct analytical capabilities. A 2025 Bristol Myers Squibb AIChE conference abstract likewise describes using experimental systems and CFD to understand reactor and purification scale-up effects on critical quality attributes.

Those sources do not establish a universal CFD-to-approval shortcut. Samsung's October study is company-reported work; a model, a measured batch result and a regulatory determination are three different forms of evidence. The BMS comparison uses only the publicly available conference abstract.

A 500 L capability also carries a containment responsibility

Samsung's corporate timeline lists its dedicated ADC facility as CGMP Ready in February 2025. Its bioconjugation service describes single-use and stainless-steel equipment supporting processes of up to 500 L. The 2023 independent study reached up to 50 L; its figures cannot be transferred as validation data for Samsung's larger installation.

Some ADC payloads are highly potent substances. Samsung says the facility's containment design meets a stringent occupational exposure limit (OEL) under 1 ng/m³. This is a company statement about facility handling and containment standards, not an independent set of measured exposure results across every worker and operating condition.

Product quality and worker safety are both essential, but their evidence and ownership differ. Process development and QC address conjugation attributes and residual material; containment engineering, safety procedures and exposure controls protect personnel. One cannot be inferred automatically from the other.

Where fluid mechanics meets ADC process science, QC and MSAT

Fluid dynamics and reactor modeling can transfer from conventional chemical engineering into biomanufacturing. But a CFD specialist does not automatically know ADC chemistry, GMP documentation or drug-specific critical quality attributes. A bioprocess specialist may also lack experience validating CFD model assumptions.

ADC process development works on conjugation conditions; analytical development and QC measure drug loading and residuals; MSAT connects development to manufacturing; engineering and safety teams manage containment. Samsung's official MSAT role feature distinguishes manufacturing technology transfer (MT), process characterization and optimization (PD), and analytics.

A hiring brief written only as 'ten years in biopharma' would miss those boundaries. More decision-relevant questions are which reaction systems an applicant modeled, what data validated the model, which analytical methods they used, and how far they personally took a process into production. These are Banseog capability inferences, not newly advertised Samsung openings.

BANSEOG VIEW | The capability to explain scale matters as much as capacity

Samsung has publicly described a CGMP Ready ADC facility, conjugation capability up to 500 L, and new CFD work on scale-up. These indicate infrastructure and technical direction. They should not be collapsed into proof that every customer ADC has completed every form of process and regulatory verification.

The 2023 study's different mixing predictions alongside limited changes in final DAR for a specific reaction reveal why this is a modeling problem rather than a 'bigger is always riskier' slogan. A company needs to connect hydrodynamics, kinetics, measured product attributes and GMP controls.

From a talent-strategy standpoint, the key is to define which person or team can justify that connection: CFD modelers, conjugation scientists, QC analysts, MSAT engineers and safety experts may hold different pieces of the evidence. The scarce capability is not merely operating a larger reactor but explaining—and demonstrating—why the desired quality should survive the change of scale.

BANSEOG VIEW

Banseog View — Model prediction, product analytics and GMP are distinct responsibilities

Samsung's up-to-500 L infrastructure and 2026 CFD announcement are company facts; mixing predictions of 9.4, 32.2 and 17.6 seconds belong to a separate 2023 study.

For a specific DAR 2 reaction, differing mixing times did not substantially alter final DAR; the relevance of mixing requires kinetics and direct analytical evidence.

Talent needs span chemical-engineering CFD, ADC conjugation, QC/analytics, MSAT manufacturing transfer and worker-safety engineering, with different qualification boundaries.

SOURCES

Primary sources and references

Research and corporate sources reviewed as of 12 October 2026. Samsung's October 7 CFD findings are company-reported; full experimental results, product-specific GMP scope and regulatory approvals are not independently established here. The 9.4 s (GST-1), 32.2 s (GST-2) and 17.6 s (SUM) results are CFD-predicted mixing times for different vessels in a separate 2023 KIT/AstraZeneca study up to 50 L, focused in part on specific DAR 2 kinetics; they are not Samsung's 500 L measurements. A June 2023 corrigendum to that paper stated its scientific conclusions were unaffected. The 2025 BMS AIChE source is a public abstract only. The OEL under 1 ng/m³ is Samsung's stated containment/handling criterion, not an independently measured exposure dataset. All job capability-transfer interpretations are Banseog analysis rather than confirmed hiring demand.