☰

Dairy Biotechnologies Sp. z o.o. (Nomi Biotech Corporation)

Scaling functional dairy fermentation from benchtop to production

Industry
Functional Dairy Biotechnology (Precision Fermentation)
Headquarters
Złotniki, Poland
Public information as of
January 2026

A4BEE prepared this analysis from publicly available sources. It reflects our own reading of Dairy Biotechnologies Sp. z o.o. (Nomi Biotech Corporation)'s published strategy and is not endorsed by, or produced in cooperation with, Dairy Biotechnologies Sp. z o.o. (Nomi Biotech Corporation). Company website

Strategic priorities

Dairy Biotechnologies Sp. z o.o. is the dairy-application arm of Nomi Biotech Corporation, a Polish-Japanese biotechnology group based at the YouNick Technology Park in Złotniki and manufacturing at Puławy, with a Taiwanese subsidiary for *Curcuma longa* rhizome processing and a stated ambition to reach seven or more country markets under the milcu brand. The commercial pipeline rests on NOMICU L-100, a patent-pending nano-emulsion technology that the company describes as making curcuminoids ten thousand times more dispersible in water, paired with *Bifidobacterium animalis* subsp. *lactis* BB-12 probiotic cultures that the brand promises at clinically validated counts.

Scaling is the central operational question. The company has guided in published coverage to fermenter volumes up to ten thousand litres, while the academic literature on its curcumin-probiotic combination documents that *B. animalis* counts must remain above seven to nine log CFU per gram for twenty-eight days even though curcumin is intrinsically antimicrobial. That tension, between an additive that tends to suppress the live organism the brand sells, runs through every batch and sets the precision bar for both the fermentation and the cleaning cycles that follow it on multi-product dairy lines.

Dairy Biotechnologies is funded through the Polish National Centre for Research and Development Fast Track programme and the Bridge Alfa seed round led by Black Pearls VC, and is working through the National Medicines Institute on what it describes as a complex novel-food certification path. With a seven-country expansion already underway in the United Kingdom, Sweden, Finland, Hungary and the United Arab Emirates, the regulatory work, the clinical-trial provenance work and the multi-site IT/OT work all sit on the same critical path, and each of them leans on the same underlying capability: data that can be read outside the equipment that produced it.

Challenges we see

  • Operations Manufacturing

    Holding dispersion stability across a ten-thousand-litre scale-up

    The NOMICU L-100 platform describes a ten-thousand-fold improvement in curcuminoid dispersibility, and the company's stated ambition is to ferment at volumes up to ten thousand litres. Published research on the curcumin-probiotic combination notes that standard turmeric extracts form insoluble precipitates and lose colour stability, so the homogeneity the benchtop process achieves has to survive a non-linear scale-up into viscous dairy matrices.

    Where dispersion quality is confirmed by finished-product sampling, the population under review is the whole batch. Reading particle distribution and emulsion state during processing narrows the population to the units that are actually deviating.

  • Operations Manufacturing

    Keeping probiotic viability stable across the antimicrobial conflict

    The fermentation has to hold *Bifidobacterium animalis* counts above seven to nine log CFU per gram for twenty-eight days of shelf life, and the literature on the same combination reports that curcumin is intrinsically antimicrobial. The brand is sold on the live-organism count, so the biological tension sits directly on the commercial proposition.

    Where probiotic viability is confirmed by plating at the end of fermentation, the corrective action arrives after the batch is sealed. Tracking viability-influencing parameters through the run lets a parameter drift be corrected before the biological window closes.

  • Data Integration

    Linking clinical-trial batches to commercial production

    Operations span three zones: Złotniki research and development, Puławy fermentation and packaging, and a Taiwanese subsidiary that processes the rhizomes. Clinical trial efficacy is generated from specific golden-batch formulations in the Złotniki lab, and there is no shared data model that connects a Taiwanese rhizome lot through a Złotniki protocol to a Puławy fermenter.

    Where each site's data sits in its own system, the chain of custody between a clinical claim and a commercial jar is reconstructed on paper. A shared model makes the chain queryable as one record rather than assembled at audit time.

  • Compliance Regulatory

    Sustaining novel-food and health-claim evidence under shifting rules

    The company is moving through a novel-food certification process with the National Medicines Institute under EU 2015/2283 and markets a product in a category where the World Health Organization has recently reclassified hangover in ICD-11 (the International Classification of Diseases, eleventh revision). Health-claim rules differ between FDA, EFSA and other regulators and continue to evolve.

    Where a claim is documented once and re-used across markets, a rule change in one jurisdiction becomes an inspection finding in another. Tying claim text to its underlying record and its target market makes the update surface visible the moment the rule moves.

  • Operations Manufacturing

    Running high-stress homogenisation without surfactant chemistry

    Achieving ten-thousand-fold dispersibility without synthetic emulsifiers relies on high-pressure homogenisation or ultrasonic cavitation, both of which put heavy mechanical stress on critical equipment. Clean-label constraints keep the formulation simple while the equipment load rises.

    Where homogeniser condition is confirmed by finished-product testing, valve drift only becomes visible after particle-size distribution has already moved. Tracking pressure, vibration and acoustic signature during the run surfaces valve wear while the equipment is still producing in-spec material.

Opportunities, by urgency and business impact

Each bubble is one opportunity, numbered to match the list below. Further right means it bites sooner; higher means a bigger effect on the business. A bigger bubble means a bigger implementation effort.

Source: A4BEE analysis of public sources

  1. Reading fermentation conditions in real time during the run

    Bifidobacterium viability is mostly confirmed by plating samples after fermentation has finished, and the published literature on the curcumin-probiotic combination shows that viability loss can occur during the run itself.

    Bringing pH, dissolved oxygen, temperature and curcumin concentration into a single process model lets a parameter drift be flagged against the batch that is currently running, and gives the operations team the same view of the process that the microbiologist has after plating.

    • Frontiers in Nutrition, 2023, on curcumin additives and probiotic viability in low-fat yogurt
    • EurekAlert, 2023, on purified curcumin preserving probiotic viability

  2. Tracing the clinical-to-commercial chain as a single record

    A clinical-trial golden batch from the Złotniki lab and a commercial batch from Puławy are described in different systems with no shared identifier, so the chain of custody between a health claim and the jar on a shelf is reconstructed by hand.

    An ontology-driven data platform that defines rhizome lot, formulation, fermentation run, packaging run and clinical reference once lets the same query return the answer from either side of the chain, instead of one reconciliation per report.

    • Nomi Biotech Corporation, corporate site, accessed January 2026
    • Nomi Biotech research collaboration disclosures with the National Medicines Institute

  3. Adjusting formulation to absorb raw-material variability

    *Curcuma longa* rhizomes from Taiwan vary in curcuminoid content and moisture profile with soil, harvest timing and climate, and a fixed extraction recipe in Poland cannot absorb that variability without finished-product drift.

    Pairing an incoming-material fingerprint with an adaptive recipe engine lets temperature, solvent ratio and mixing time be calibrated per lot, so output potency is held to specification even when input variability is high.

    • Deep research note on rhizome sourcing variability
    • NOMICU L-100 product literature on clean-label extraction

  4. Predicting homogeniser wear before it changes the particle size

    High-pressure homogenisers are single points of failure on the line, and finished-product testing only detects particle-size drift after a batch has already been produced.

    Continuous vibration and acoustic sensing on the homogeniser builds a wear model per machine, so a planned maintenance window is scheduled around production rather than forced by a particle-size excursion.

    • Industry coverage of clean-label processing constraints
    • Deep research note on equipment load in nano-emulsion production

  5. Keeping multi-market labels and claims under one source of truth

    A claim that is compliant in one of the seven target markets can be non-compliant in another, and manual monitoring of regulatory feeds cannot keep pace with FDA, EFSA and FSANZ updates across a seven-country footprint.

    Tying label text and health claims to a structured regulatory layer that knows the market, the claim category and the latest guidance turns a label change from a manual exercise into a controlled update with a named approver.

    • Proceedings of the 13th Alcohol Hangover Research Group Meeting, MDPI
    • Nomi Biotech market expansion disclosures

What we'd propose

  • Digital Lab

    Real-time fermentation intelligence for probiotic dairy runs

    We instrument the fermenters, bring pH, dissolved oxygen, temperature and curcumin concentration into one process model, and run a viability proxy against the batch that is currently running, so a parameter drift is visible to operations while the run is still recoverable.

    • Fermenter data acquisition

      Getting data off the vessel

      Connect bioreactor controllers and side-mounted sensors through OPC UA (Open Platform Communications Unified Architecture — a vendor-neutral machine-to-machine protocol) or MQTT (a lightweight messaging protocol commonly used for industrial telemetry) so process values leave the vessel in a documented form rather than staying inside the controller.

    • Viability proxy against the running batch

      A drift signal during the run

      Build a viability-influencing model from historical batches where plating results are available, then run the same model on the live batch every few minutes, so operators see a drift signal long before the final plating arrives.

    • Run record assembled as data

      Evidence generated by the line

      Write every monitored parameter into a per-batch record with lineage back to the instrument that produced each value, so the evidence behind a release decision is generated by the fermenter rather than compiled from it.

    • Parameter drift is visible to operations during the run, not after plating.
    • The same process model serves operations, quality and the academic partners instead of three separate extracts.
    • The release record carries the data the inspection would have asked for.

    We've done this before

    See how we do Digital Lab
  • Enterprise AI

    Ontology-driven data platform across Taiwan, Złotniki and Puławy

    An ontology-based data platform that defines rhizome lot, formulation, fermentation run, packaging run and clinical reference once, then loads the Taiwanese subsidiary, the Złotniki lab and the Puławy production site against that single model so the clinical-to-commercial chain becomes queryable.

    • Shared product and batch ontology

      One agreed set of terms

      Define rhizome lot, formulation version, fermentation run, packaging run and clinical reference as explicit entities with agreed relationships, so a query written once returns comparable answers from all three sites instead of three dialects of the same table.

    • Pipelines from LIMS, MES and ERP

      Loading all three sites

      Build ingestion for the Złotniki LIMS (Laboratory Information Management System), the Puławy MES (Manufacturing Execution System) and the Taiwanese subsidiary's batch records, with schema validation at the boundary so a malformed record fails loudly rather than silently.

    • Retrieval and reporting on top of the model

      Questions answered without IT tickets

      Expose the model through dashboards and a retrieval layer so quality, regulatory and clinical teams can ask questions of the combined dataset without commissioning a new extract for each one.

    • Integration work is done once against a shared model instead of once per point-to-point interface.
    • The chain of custody between a clinical claim and a commercial jar is queryable rather than reconstructed.
    • New sites and new assays attach to the model rather than triggering another migration.

    We've done this before

    See how we do Enterprise AI
  • Digital Lab

    Adaptive formulation for variable rhizome inputs

    A spectroscopic fingerprint on every incoming rhizome lot feeds an adaptive recipe engine that adjusts temperature, solvent ratio and mixing time per lot, so output potency is held to specification when input variability is high.

    • Incoming-material fingerprint

      A reading on every lot

      Near-infrared or Raman spectroscopy on each incoming rhizome shipment produces a per-lot curcuminoid and moisture profile in minutes, with the result captured as data rather than entered by hand.

    • Adaptive recipe calculation

      Parameters set per lot

      A recipe engine uses the lot fingerprint to set extraction temperature, solvent ratio and mixing time for that lot, so the static recipe becomes a per-lot calibrated recipe without changing the equipment.

    • Output consistency check

      Closing the loop

      An output-side check compares the actual finished-product potency to the target band, and the deviation feeds back into the recipe engine so the next lot's calibration improves rather than the next lot's failure repeats.

    • Output potency stays in band even when rhizome variability is high.
    • Raw-material waste from recipe failure drops, and supplier flexibility rises.
    • The fingerprint and the recipe together become a defensible specification for the clinical partners.

    We've done this before

    See how we do Digital Lab
  • Digital CDMO

    Condition-based maintenance for high-pressure homogenisers

    Vibration and acoustic sensors on the high-pressure homogenisers feed a per-machine wear model, so valve replacement is scheduled around production rather than forced by a particle-size excursion.

    • Vibration and acoustic sensing

      Sensors on the critical machine

      Mount vibration and acoustic sensors on the homogeniser valve stack, with acquisition through OPC UA so the signal reaches the platform in a documented form rather than being logged locally.

    • Per-machine wear model

      Predicting valve drift

      Build a wear model per machine from historical runs, so the model flags when a valve is approaching the particle-size-drift threshold rather than after the threshold has been crossed.

    • Maintenance window recommendation

      Planned downtime, not forced downtime

      Combine the wear model with the production schedule to recommend a maintenance window that fits the next planned downtime, with the option to hold a batch when the model is signalling a high-risk run.

    • Valve replacement is scheduled, not forced by a particle-size excursion.
    • Equipment data feeds the same platform the fermentation intelligence proposal uses.
    • Maintenance team spends time on planned work rather than on unplanned intervention.

    We've done this before

    See how we do Digital CDMO
  • Agents

    AI agents for novel-food and health-claim documentation

    Narrow, reviewable agents that take the repetitive part of regulatory work: drafting novel-food dossier sections from source records, checking a label against its target market before it enters review, and finding every controlled document a regulatory change affects. A named person approves every output.

    • Drafting from source records

      First drafts from system data

      Generate the first draft of a novel-food section, a clinical-claim justification or a regulatory update summary directly from the underlying LIMS, MES and clinical records, so the author edits and judges rather than assembles.

    • Template and market-rule check

      Gaps found before review

      Check a submitted label or dossier section against the template for its target market (FDA, EFSA, FSANZ), returning missing or inconsistent sections before it enters the human review queue.

    • Change impact search across the controlled document set

      Which documents a rule change touches

      When a regulator updates guidance, retrieve every controlled document, label and claim text that references it and rank them by how directly they are affected, so the update scope is known on day one.

    • Review queues move faster because documents arrive complete.
    • The scope of a regulatory change is established by search rather than by recollection.
    • Every output is traceable to the source records it came from and signed off by a named reviewer.

    We've done this before

    See how we do Agents

Where QB Systems fits

Alongside our services we build QB Systems, hardware and software for bioprocess control. QB Systems is a product brand of A4BEE Sp. z o.o.

Applications
Bioreactors

Software-defined control for a bioreactor — a new QB vessel, an upgrade to one you have, or a retrofit of the existing PLC.

Buffer & media preparation

Automated preparation of growth media and process buffers, so a recipe runs the same way every time without fixed infrastructure.

Automated sampling

Automated sampling from 4–18 sources, aseptic-capable and up to 72 hours unattended. Works with any vendor's bioreactor.

Deployment
Retrofit

Existing equipment keeps running; QB takes over the PLC, or reads from it without touching control.

Scale
Pilot (50–300 L)

Stainless steel, where QB supplies the control software and integration and a certified partner builds the installation.

Digital maturity: today and target

Scored out of 100 across six dimensions. The target is what Dairy Biotechnologies Sp. z o.o. (Nomi Biotech Corporation)'s own published ambition implies — not a perfect score.

Source: A4BEE analysis of public sources
Lab-to-line data 32 → 78
Three operational zones (Złotniki, Puławy, Taiwan) carry separate data estates and the published collaboration with the National Medicines Institute adds a fourth academic shape, so a clinical-to-commercial query currently requires manual reconciliation.
Process monitoring 38 → 82
Fermentation monitoring runs on periodic sampling and finished-product testing rather than on continuous digital oversight, which fits a Seed-stage operation but constrains what can be done at the ten-thousand-litre scale.
Predictive analytics 22 → 70
Batch outcomes are confirmed after production rather than predicted during the run, so corrective action arrives after the biological window has closed; the platform work to change this has not yet started.
IT/OT convergence 28 → 78
Laboratory instruments, fermenters and homogenisers operate as isolated data islands, so real-time correlation between formulation specification and fermentation execution is not yet possible.
Regulatory document systems 33 → 72
NCBR (Polish National Centre for Research and Development) grant reporting and novel-food submissions are compiled manually from disconnected sources, placing the burden on the same scientific talent that runs the experiments.
Supply chain traceability 30 → 70
The Taiwan-to-Puławy rhizome flow is tracked through fragmented systems, with quality attributes lost between handoff points and no end-to-end query path from rhizome lot to finished jar.

Check this yourself

Our Service Portal has free self-assessments and market comparisons. These are the ones that line up with what we've read above — no sales call required.

Think we've read this right?

Talk to us

Related reading

This is an independent analysis prepared by A4BEE from publicly available information as of January 2026. It reflects A4BEE's own interpretation and opinion, is not affiliated with, endorsed by, or verified with Dairy Biotechnologies Sp. z o.o. (Nomi Biotech Corporation), and may be incomplete or inaccurate. All company names and trademarks are the property of their respective owners. To request a correction or removal, contact [email protected].