☰

ConvEyXO

Portable exosome manufacturing at clinical scale

Industry
Biotechnology
Headquarters
Gosselies, Belgium
Public information as of
May 2026

A4BEE prepared this analysis from publicly available sources. It reflects our own reading of ConvEyXO's published strategy and is not endorsed by, or produced in cooperation with, ConvEyXO. Company website

Strategic priorities

ConvEyXO is preparing to enter the clinic in 2027 with two RNA-enhanced exosome candidates, EXOSTEO01 and DERMEXO01, and is industrialising a proprietary 3D-printed fixed-bed bioreactor platform and a microfluidic mechanoporation loading technology developed with its XomeXBio joint venture. The company was founded in 2019 and operates from Biopark Charleroi South in Gosselies, Belgium, with its 2027 milestone dependent on transferring these bespoke processes into a CDMO environment such as the Catalent campus at Gosselies.

Total funding to date stands at roughly 6.45 million US dollars across a July 2020 early-stage round and an August 2023 seed round led by SambrInvest, and the company is already generating revenue from its Exo-Harvest affiliate. Capital efficiency is a stated operating principle, so the digital work that scales the process has to do so on the same hardware footprint rather than asking for a step-change in headcount or capex.

Three layers have to come together for the 2027 milestone: process data leaving the 3D bioreactor and microfluidic rig in a standardised form so the platform can be redeployed at a CDMO without rewriting orchestration logic; analytical and process data combined into a single model that supports regulatory submissions for a modality that does not yet have agreed reference standards; and operational records that meet ALCOA+ data integrity expectations as the company moves from R&D into GMP operations.

All three lean on the same underlying capability: process and quality data that can be read by systems outside the equipment that produced it, both at ConvEyXO's own Gosselies sites and at the CDMO the technology is transferred to. That is the practical meaning of a 'Plug & Produce' tech-transfer backbone in a company whose competitive position rests on hardware that no off-the-shelf controller natively understands.

Challenges we see

  • Operations Manufacturing

    Reading oxygen and shear gradients inside 3D-printed scaffolds

    ConvEyXO's 3D-printed fixed-bed scaffolds create complex micro-environments in which oxygen and nutrient flow can be non-homogeneous, generating hypoxic zones that degrade exosome potency. The company describes the resulting yield and consistency as a major constraint on moving from 24-condition mini-bioreactors to commercial-scale reactors.

    Where perfusion behaviour is inferred from the outlet of a fixed-bed reactor, the inside of the scaffold is effectively inaccessible during a run. Bringing the spatial profile into a measured or modelled view narrows the gap between the design geometry and the conditions the cells actually experience.

  • Compliance Operations

    Moving batch records off spreadsheets ahead of GMP operations

    Senior Cell Culture Technicians manage batch records, traceability and SOPs in Microsoft Excel and Word, an arrangement that becomes harder to defend under ALCOA+ data integrity expectations as ConvEyXO transitions toward GMP operations for the 2027 clinical entry.

    When a batch record is composed in a general-purpose office tool, every step is a separate manual action to perform and to verify, and the audit trail lives alongside the data rather than inside it. Carrying the same workflow into a regulated environment means changing the medium as much as the procedure.

  • Compliance Regulatory

    Making a bespoke reactor redeployable inside a CDMO cleanroom

    The Exo-Harvest 3D bioreactor and the XomeXBio mechanoporation rig are proprietary DeepTech assets that do not natively fit Catalent's standardised GMP cleanrooms, and the Catalent Gosselies campus is the natural first destination for clinical-scale production. Transferring a non-standard process to a standard cleanroom is slow and expensive without a process-level abstraction.

    Where the orchestration logic for a reactor is written against its specific controller, the same process has to be rewritten at every new site. Exposing the reactor through a standard module description means the orchestration moves with the process rather than being rebuilt at each transfer.

  • Digital Integration

    Joining analytical and process data into one working model

    Exosome characterisation results from Nanosight, Western Blot and total protein assays are kept in spreadsheets disconnected from bioreactor SCADA streams, and an internal AI tool is used for scaffold geometry selection but is not integrated with the production telemetry. The result is the inability to assemble a Golden Batch profile that links growth conditions to therapeutic potency.

    When the design tool, the SCADA stream and the offline characterisation live in separate places, every correlation between input and potency is rebuilt by hand. Aligning them under one schema turns those correlations into something the team can query rather than reconstruct.

  • Digital Regulatory

    Hardening OT cybersecurity as operations become more closed

    As Exo-Harvest moves toward closed automated operation, the OT layer still lacks IEC 62443-aligned segmentation, and the prevailing tooling posture remains oriented around generic Microsoft Office software rather than industrial-grade control systems. NIS2 obligations apply as operations close.

    As more of the process moves behind closed automated control, the network carrying it becomes part of the regulated system. Treating that network as an industrial control network from the start means the security posture does not have to be re-argued later against a live production line.

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. Modelling oxygen and shear conditions across the 3D scaffold

    ConvEyXO cannot resolve oxygen and shear-stress gradients within its 3D-printed fixed-bed reactors, so the scaffolds are physical black boxes that limit yield and consistency.

    A spatial digital twin that combines CFD models of the scaffold geometry with edge-AI sensor inference predicts dissolved oxygen across the bed and adjusts perfusion rate in real time, keeping conditions within a defined envelope without requiring constant operator attention.

    • ConvEyXO Manufacturing: Closed Automated Operation whitepaper
    • ConvEyXO Manufacturing page, AI tool for scaffold geometry selection

  2. Putting batch records on a 21 CFR Part 11 platform

    Day-to-day cell culture, traceability and batch documentation are run through Excel spreadsheets and PowerPoint, creating an unsustainable manual burden and an ALCOA+ audit risk as the company transitions to GMP.

    A GAMP5-compliant Laboratory Execution System captures instrument data automatically, enforces electronic signatures and aligns SOPs across the Gosselies sites, so the same record format follows a batch from research through to clinical release.

    • ConvEyXO Senior Cell Culture Technician job posting
    • XomeXBio joint venture press release, SomesTech collaboration

  3. Wrapping Exo-Harvest and XomeXBio as MTP modules

    The proprietary Exo-Harvest and XomeXBio platforms are not standardised for Plug & Produce deployment at Catalent or other CDMOs, which puts the 2027 clinical schedule at risk during tech transfer.

    Implementing MTP (Module Type Package) and OPC UA wrappers around the bioreactor and mechanoporation hardware exposes them as standardised modules that can be orchestrated by any compliant process orchestration layer, shortening onboarding at each new site.

    • Catalent Gosselies campus page
    • XomeXBio joint venture press release on instrumentation validation

  4. Unifying CAD, SCADA and analytical data into one model

    Offline analytical results (particle concentration, miRNA expression, total protein assays) are siloed from online sensor data, which prevents identification of the Golden Batch and slows regulatory documentation.

    An ontology-based industrial data platform unifies CAD scaffold designs, SCADA streams and offline analytics into a single source of truth that feeds multivariate yield models, so the Golden Batch profile is built once and reused for every new run.

    • MDPI Pharmaceutics review on exosome reference standards
    • ConvEyXO Manufacturing page, AI tool for scaffold geometry selection

  5. Migrating microfluidic and AI workloads to a resilient compute layer

    Microfluidic simulation and AI geometry-selection workflows run on isolated workstations, creating single points of failure that jeopardise design-thread continuity from CAD to bioreactor.

    A high-availability Incus/LXC cluster hosts simulation and AI workloads in containers with automatic failover, so a hardware fault does not halt the AI geometry tool or the simulations that feed into new scaffold designs.

    • Hui Yang bio, Shenzhen Institute of Advanced Technology
    • XomeXBio joint venture press release, IPRATECH partnership

What we'd propose

  • Digital CDMO

    Spatial digital twin for the 3D fixed-bed bioreactor

    A closed-loop digital twin of the Exo-Harvest scaffold that combines CFD models, real-time sensor data and edge-AI control to keep oxygen and shear conditions inside a defined envelope across the bed.

    • CFD-driven twin engine

      Spatial oxygen and shear prediction

      Couple 3D scaffold CAD geometries with computational fluid dynamics to predict oxygen and nutrient gradients across the perfusion bed at industrial scale, so the same model used to design the scaffold also describes the conditions inside it during a run.

    • Edge-AI perfusion control

      Real-time adaptive flow tuning

      Run lightweight ML inference at the reactor edge to adjust perfusion rate against the predicted gradient, so hypoxic clusters are addressed before they affect exosome potency rather than being investigated after a batch release.

    • Synthetic data generator

      Simulation-driven dry runs

      Produce high-fidelity synthetic batches that train control models without consuming physical media, so new cell lines and cargoes can be characterised against the twin before they reach the bench.

    • Hypoxic zones are predicted and prevented during the run rather than inferred from a yield deviation afterwards.
    • The same CFD model is shared between scaffold design and process control, so the geometry that wins the simulation is also the one the control system knows about.
    • New cell lines and cargoes are qualified against the twin before physical runs, shortening R&D cycles for new candidates.

    We've done this before

    See how we do Digital CDMO
  • Digital Lab

    GAMP5 Laboratory Execution System for the Gosselies sites

    A 21 CFR Part 11-compliant LES that replaces Excel-based batch records, automates instrument data capture and aligns SOPs across ConvEyXO, Exo-Harvest and XomeXBio.

    • Automated data capture

      Direct instrument-to-LES streaming

      Connect Nanosight, balances, pH meters and other QC instruments to the LES via barcode and OPC UA so results carry instrument identity, method version and timestamp, removing the manual transcription step that currently sits in the record path.

    • Harmonised digital workflows

      Unified pH, balance, conductivity steps

      Enforce a single digital SOP across the Gosselies sites so the same procedure is followed at ConvEyXO, Exo-Harvest and XomeXBio, and deviations are recorded the same way regardless of where the batch was prepared.

    • Audit-ready electronic records

      Inspection-grade evidence

      Generate timestamped, attributable batch records with electronic signatures that meet ALCOA+ expectations, so an FDA or EMA inspection can be answered from the system rather than from reconstructed paperwork.

    • Manual transcription between instruments, spreadsheets and the batch record is removed from the release path.
    • An SOP change is rolled out once across all Gosselies sites instead of being chased through individual Excel templates.
    • Audit questions are answered from the record itself rather than from a reconstruction.

    We've done this before

    See how we do Digital Lab
  • Enterprise AI

    MTP-based tech-transfer enablement for the 2027 milestone

    Module Type Package and OPC UA standardisation of the Exo-Harvest reactor and XomeXBio mechanoporation rigs, so the process can be redeployed at a CDMO cleanroom without rewriting orchestration logic.

    • MTP module wrapping

      Hardware abstraction for portability

      Encapsulate the proprietary 3D bioreactor and microfluidic loader as MTP-compliant modules, so any MTP-aware process orchestration layer can drive them at ConvEyXO, at Exo-Harvest and at a CDMO without custom integration work.

    • OPC UA backbone

      Secure machine-to-machine communication

      Replace ad-hoc protocols with OPC UA so process parameters and CQAs are recorded with millisecond precision across IT and OT systems, giving the validation team one data path to evidence regardless of which site the batch was made at.

    • Tech-transfer readiness audit

      Catalent-grade gap assessment

      Benchmark ConvEyXO's process maturity against Catalent Gosselies' GMP requirements and produce a remediation roadmap for the 2027 milestone, so the gap between the current process description and a CDMO-acceptable one is known before transfer is scheduled.

    • Onboarding time at Catalent or another CDMO drops because the orchestration logic travels with the process.
    • Proprietary IP stays protected inside a documented module description rather than being exposed in custom integration code.
    • Future sites replicate the same description rather than rebuilding it.

    We've done this before

    See how we do Enterprise AI
  • Enterprise AI

    Industrial data platform for exosome characterisation

    An ontology-based platform that unifies CAD, SCADA and offline analytics into a single Golden Batch model for exosome therapeutics.

    • Ontology-based data model

      Unified biotech taxonomy

      Model exosome attributes, scaffold geometries and process parameters in a single semantic graph, so multivariate models correlate design inputs to potency outcomes without each new analysis rebuilding the data set from scratch.

    • Golden batch overlay

      Live deviation visualisation

      Compare running batches against the historical optimum in Grafana, surfacing deviations in CSPR, VCD and particle concentration as they develop, so the operations team sees the same view of a batch that the analytical team will see after release.

    • PAT integration layer

      Offline-online data fusion

      Bridge low-frequency Nanosight and Western Blot results with high-frequency reactor sensors so Quality-by-Control decisions are made on combined data rather than whichever stream happened to be available at the moment of review.

    • The Golden Batch profile is built once and reused, instead of being rebuilt by hand for every regulatory submission.
    • The AI scaffold geometry tool becomes a licensable digital asset attached to a working data layer rather than a standalone proof of concept.
    • Regulators see one combined evidence trail rather than reconciled extracts from separate systems.

    We've done this before

    See how we do Enterprise AI
  • Digital Lab

    High-availability R&D compute cluster for the Gosselies hubs

    An Incus/LXC-based high-availability compute cluster that hosts microfluidic simulations, AI geometry tools and digital lab applications across the Gosselies hubs.

    • Incus/LXC orchestration

      Container-native lab compute

      Encapsulate simulation and AI workloads in lightweight containers so deployment between R&D and pre-GMP environments is consistent and reversible, removing the drift that appears when a tool is rebuilt by hand at each new site.

    • HA cluster failover

      No-single-point-of-failure design

      Migrate workloads automatically between cluster nodes so a hardware failure does not halt microfluidic simulations or AI runs, keeping the design thread from CAD to bioreactor continuous across overnight and weekend runs.

    • Infrastructure-as-Code site templates

      Reproducible global deployment

      Provide IaC blueprints so XomeXBio and Exo-Harvest sites stay configuration-aligned as the company scales, and so a new affiliate or CDMO partner can be brought onto the same compute environment with a known starting point.

    • A hardware fault no longer halts the AI geometry tool or the simulations that feed scaffold design.
    • R&D compute stays configuration-aligned across ConvEyXO, Exo-Harvest and XomeXBio as new sites are added.
    • Reproducible container environments shorten the path from a simulation result to a physical batch.

    We've done this before

    See how we do Digital Lab

Digital maturity: today and target

Scored out of 100 across six dimensions. The target is what ConvEyXO's own published ambition implies — not a perfect score.

Source: A4BEE analysis of public sources
Process automation 30 → 85
Operations rely on manual cell-culture technicians and Excel-based traceability; closed automated operation is the stated direction, and the digital infrastructure to support it is still being assembled.
Data integration 25 → 85
Analytical, design and SCADA data are siloed; the AI geometry tool is not integrated with production telemetry, so correlations between inputs and potency are reconstructed by hand.
Regulatory and compliance readiness 35 → 90
Paper-based records and the absence of ALCOA+ tooling create exposure ahead of the 2027 clinical milestone; the work to bring records onto a 21 CFR Part 11 platform is the practical prerequisite for inspection-grade submissions.
Tech-transfer portability 30 → 80
Proprietary 3D reactors and mechanoporation rigs are not yet MTP- or OPC UA-standardised, so each transfer to a CDMO carries the cost of bespoke integration work.
OT cybersecurity 25 → 80
The OT layer lacks IEC 62443-aligned segmentation; as operations close, the network carrying the process becomes part of the regulated system and has to be treated as such from the start.
Advanced analytics and AI 45 → 85
An internal AI tool exists for scaffold geometry selection but is not yet productised into a digital twin or a Quality-by-Control loop; the data layer it needs is the work in progress.

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 May 2026. It reflects A4BEE's own interpretation and opinion, is not affiliated with, endorsed by, or verified with ConvEyXO, 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].