☰

Astraveus

One benchtop factory, replicated at point of care

Industry
Cell and Gene Therapy Manufacturing
Headquarters
Paris, Île-de-France, France
Public information as of
January 2026

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

Strategic priorities

Astraveus has built the Lakhesys Benchtop Cell Factory, a microfluidic platform that compresses the footprint of a roughly 400 square metre cell-therapy facility into a six-foot benchtop unit and that demonstrated its first end-to-end autologous CAR-T production in early 2025. The company is now moving from a single functional prototype to an industrial fleet of modular foundries capable of processing up to 10 patient doses simultaneously inside one closed system. Founding work at Saint-Louis Hospital in 2016 and over €30 million of combined private and public funding (a €16.5 million seed round, a €10.4 million France 2030 grant in 2023 and a €7.1 million Première usine grant in 2024) frame the scale of the industrialization task.

The economic case the platform rests on is explicit: per-dose CAR-T manufacturing costs are targeted to fall by a factor of five, and vein-to-vein time from a typical three to six weeks down to about 26 hours. Both depend on moving the platform out of a single development laboratory into multiple lower-classification cleanrooms close to patients, with onboard analytics replacing the offline flow cytometry and manual microscopy that central facilities rely on. Didier Masson joined as Chief Operating Officer in 2024 and Ken Kotz as Chief Technology Officer in 2025, signalling that the next phase is operations and industrial engineering rather than further biological proof of concept.

The most concrete near-term pull is a 'Premiere usine' microfluidic consumables facility. Lines built by hand in a development laboratory have to become a factory production process, and the same Lakhesys device has to perform equivalently from one site to the next without a centralised plant manager in the loop. The benchmark is published: a 26-hour vein-to-vein target, a 5x cost reduction, and a ten-dose parallel run inside a single closed benchtop system. Every digital decision made in 2026 is being measured against that target.

These numbers all sit on three shared capabilities: a process and analytics model that runs identically on every Lakhesys unit, an integration and orchestration layer that lets a fleet of benchtop factories be supervised together, and a regulatory and quality story that is assembled continuously from the device itself rather than compiled for each site visit. None of these is purely a software choice; all of them decide whether the platform becomes a standard of care in a national health system or stays a laboratory demonstration.

Challenges we see

  • Economics Operations

    Breaking the economics of centralised cell-therapy logistics

    Per-dose CAR-T manufacturing currently sits between $100,000 and $2 million, and centralised production produces vein-to-vein times of three to six weeks that depend on cryopreservation and global shipping. Astraveus is targeting a fivefold cost reduction and a 26-hour vein-to-vein time on the Lakhesys platform.

    Where therapy production is tied to a single large facility and a frozen-ship-thaw cycle, the population under review is the entire eligible patient cohort waiting on a slot, while the units that actually deviate are the individual benchtop runs. Operating at point of care re-scopes that question to the run currently in production and to the patient whose cells it carries.

  • Engineering Integration

    Converging microfluidics, automation and cell biology on a single device

    The Lakhesys platform integrates microfluidics, mechanical and electrical engineering, automation, cell biology and surface chemistry inside a startup environment. The biological computer vision the company is built around treats each of those layers as a control system, not as a support function.

    Where disciplines meet at a single benchtop unit, the practical unit of integration becomes the device rather than the project. Decisions about interfaces, data formats and ownership are taken once per device revision and inherited by every deployment that follows.

  • Scale-up Manufacturing

    Moving microfluidic consumables from hand-built prototypes to factory production

    Following the first end-to-end CAR-T production in early 2025, Astraveus is standing up a dedicated consumables facility through the Premiere usine grant awarded in 2024. The consumables that previously left a development laboratory now need to leave a factory line at consistent quality.

    Where microfluidic consumables were previously assembled by hand, the population under review shifts from a development batch to every sterile cartridge the factory ships. Quality produced on the line rather than confirmed afterwards becomes the way to keep that population inside the specification.

  • Compliance Regulatory

    Harmonising regulatory evidence across decentralised point-of-care units

    Point-of-care units must produce results identical in quality to a centralised facility while operating under GxP and GMP in different jurisdictions. There is no established industry standard for microfluidic-based bioproduction, so onboard analytical capability has to be the substrate of regulatory acceptance.

    Where evidence is generated in many sites rather than one, the question a regulator asks is no longer 'what did this batch do' but 'what did this benchtop unit run prove', and the answer needs to be assembled from the device rather than reconstructed after the fact.

  • Workforce Operations

    Operating advanced cell therapy without PhD-level operators at every site

    The biotech sector faces a structural shortage of cell-therapy specialists and high dissatisfaction with the manual workload current manufacturing methods impose. Astraveus is targeting a 30 percent reduction in manual errors at the operator interface alongside decentralised deployment.

    Where the operator profile narrows while the device stays sophisticated, the practical surface is the interface rather than the biology. A run that an experienced PhD can supervise by hand has to become a sequence of decisions and confirmations that a trained operator can execute under supervision.

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. Driving the per-dose cost of autologous CAR-T down by a factor of five

    Traditional CAR-T manufacturing depends on large cleanrooms, bulk reagents and concentrated viral vectors, with per-dose costs sitting between $100,000 and $2 million. The current process does not capture the efficiencies of scale seen in other high-technology industries.

    Working at microfluidic scale with a closed benchtop system lets reagent use, energy and cleanroom footprint all fall together, opening a credible path to a fivefold reduction in manufacturing cost per dose on the same Lakhesys hardware.

    • Astraveus, 'SKALE' project materials and France 2030 grant award, 2023–2024
    • Astraveus strategic research, CBO commentary on CAR-T cost reduction target, January 2026

  2. Bringing vein-to-vein time down from weeks to a day

    Patients with aggressive malignancies cannot tolerate the three to six week vein-to-vein interval of central production, and the frozen-ship-thaw cycle is itself detrimental to cell quality. Treatment slot availability caps how many patients receive a therapy even after regulatory approval.

    End-to-end production in roughly 26 hours inside a benchtop unit moves the time-critical step into the same facility where the patient is treated, so the bottleneck shifts from shipping logistics to the run itself and is observable in real time.

    • Astraveus end-to-end CAR-T production announcement, 2025
    • Prof. Jérôme Larghero commentary on bead-free microfluidic selection, Astraveus research dossier

  3. Operating a fleet of benchtop factories under shared supervision

    Scaling centralised cell-therapy plants is capital-intensive and limited by the throughput of a single building. A fleet approach requires the same Lakhesys unit to deliver equivalent results in every lower-classification cleanroom it occupies, without a central plant manager on site.

    Putting the benchtop unit under a shared orchestration layer lets each run be supervised remotely, performance compared across sites, and decisions about patient slot allocation taken on actual run state rather than on the capacity declaration of a central building.

    • Astraveus SKALE project thesis on decentralised point-of-care manufacturing, 2024
    • Astraveus C-suite appointments (Masson, Kotz) and industrialisation commentary, 2024–2025

  4. Replacing offline flow cytometry with onboard analytics on the device

    Manual enrichment, transduction and monitoring steps remain prone to contamination and human error in current cell-therapy production. Confirming quality by sending samples to an offline flow cytometer introduces both delay and a separate handling step.

    Onboard fluorescence and brightfield imaging tied directly into the device's control loop lets every batch be classified by the same hardware that produced it, with the decision record generated as part of the run rather than assembled from separate laboratory reads.

    • Astraveus research dossier on integrated onboard analytical capability
    • Lars Peeck commentary on full automation of the biological computer, Astraveus research dossier

  5. Keeping one process development workflow from preclinical to commercial

    Switching technologies between preclinical, clinical and commercial stages introduces variability and forces revalidation of comparable process steps. The same therapy concept can behave differently depending on which device produced the cells.

    Reusing identical Lakhesys bioprocessors across all development stages holds the device constant, so the variability to manage is the patient's biology rather than the platform beneath it, and validation runs do not have to be redone on a different system at every stage.

    • Astraveus strategic commentary on technology continuity across development stages

What we'd propose

  • Digital CDMO

    Module Type Package control application for each Lakhesys unit

    A vendor-neutral control application built on the Module Type Package (MTP) standard so that every Lakhesys benchtop unit exposes a uniform interface to whatever orchestration system sits above it, regardless of the underlying vendor controllers.

    • MTP service interface per module

      One description per unit

      Wrap the controllers and instruments inside a Lakhesys unit as MTP services that describe their capabilities, state and available commands, so the same orchestration system can talk to any unit that meets the same standard.

    • Plug and Produce integration

      Adding a unit without rewiring

      Add or remove a Lakhesys unit from the fleet without recoding the orchestration layer, using the MTP contract as the integration point and validating each new unit against the interface rather than against a custom configuration.

    • Bridging OPC UA, MQTT and laboratory protocols

      Reaching into instruments and sensors

      Bridge OPC UA (Open Platform Communications Unified Architecture), MQTT (a lightweight publish–subscribe messaging protocol) and the laboratory data formats already in use so the MTP service can speak to whatever instrument or sensor the next unit adds, without losing traceability of the underlying reading.

    • A new Lakhesys unit arrives with an integration story rather than as a project.
    • The fleet behaves as a fleet, so a slot at site A is comparable to a slot at site B.
    • Instrument and sensor choices stay open as the platform evolves, because the integration surface is the MTP contract.

    We've done this before

    See how we do Digital CDMO
  • Enterprise AI

    Biological digital twin and ontology for the Lakhesys run

    An ontology-based data model and a process twin of the Lakhesys run that lets Astraveus simulate, monitor and compare every run against a shared understanding of the device, the cells and the protocol.

    • Shared process ontology

      Same terms for every run

      Define run, unit, batch, sample, protocol and result as explicit entities with agreed relationships, so that data from one Lakhesys unit can be compared, queried and modelled together with data from another without a reconciliation step in between.

    • Process twin for simulation and monitoring

      A living model of the unit

      Build a process twin of the Lakhesys unit that ingests the run's time-series signals and produces a comparable view of expected behaviour, so that deviations surface against the run that is currently underway and optimisation is grounded in measured behaviour rather than in offline approximation.

    • Analytics and retrieval across the fleet

      Questions answered without new extracts

      Expose the combined model through dashboards and a retrieval layer so that medical, process development and operations teams can ask fleet-level questions without commissioning a new extract for each one.

    • One place defines what a Lakhesys run is, so every site starts from the same baseline.
    • Deviations are detected on the running batch, not reconstructed from records after release.
    • Predictive work is grounded in the same data the production line is using, which keeps simulation and operation coherent.

    We've done this before

    See how we do Enterprise AI
  • Digital Lab

    Onboard analytics and laboratory integration on the device

    An instrumentation and integration layer that turns the Lakhesys unit's onboard sensors — fluorescence, brightfield and process measurements — into a continuous data flow that reaches the laboratory and batch record systems as data rather than as observation.

    • Onboard image and process capture

      Capture at the device

      Capture fluorescence and brightfield images and process values directly from the Lakhesys unit's onboard sensors, with instrument identity, method and timestamp attached at the point of capture rather than transcribed from a screen afterwards.

    • Onboard classification pipeline

      A run classified as it runs

      Run machine-learning classification directly on the captured images to recognise cell state and process events, so that the device classifies its own run and the operator interface reflects what the device sees rather than what an offline laboratory later confirms.

    • Laboratory and batch record integration

      From device to record

      Connect the device's data flow to LIMS (Laboratory Information Management System), ELN (Electronic Lab Notebook) and batch record systems so the run's evidence is generated as part of production, with lineage back to the run that produced it.

    • Fewer manual transfers between device and laboratory, and fewer of them to verify.
    • Release decisions are made from data the device produced itself, not from a separate laboratory read.
    • A 30 percent reduction in manual errors at the operator interface is tracked as the interface is used, not estimated afterwards.

    We've done this before

    See how we do Digital Lab
  • Agents

    AI agents for protocol, regulatory and operator-facing work

    Narrow, reviewable agents that take repetitive knowledge work off the bench and the quality organisation: protocol drafting, template and completeness checking of regulatory submissions, and operator-facing onboarding tied to the actual run.

    • Protocol and study document drafting from source records

      First drafts from system data

      Generate first drafts of process descriptions, batch record narratives and protocol sections directly from the underlying run and laboratory records, so the author edits and judges rather than assembles text from scratch.

    • Regulatory and template completeness checks

      Gaps found before review

      Check a submitted document against the submission template and the site's own checklist, returning missing or inconsistent sections before the document enters the human review queue, with every output traced to the source records it came from.

    • Operator-facing run companion

      Guidance on the run that is running

      Provide an operator-facing companion that surfaces the right step, the right confirmation and the right deviation flag at the moment they are needed, anchored in the live run data rather than in a static procedure, with every recommendation traceable to a named protocol reference.

    • Submission review queues move faster because documents arrive complete and supported by links to source records.
    • Operator work shifts from manual procedure lookup to supervised confirmation, with the device's own data behind each step.
    • Continuity from preclinical to commercial is held by a single reviewable workflow instead of by institutional memory at each stage.

    We've done this before

    See how we do Agents

Digital maturity: today and target

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

Source: A4BEE analysis of public sources
Manufacturing orchestration 28 → 88
A single functional Lakhesys prototype demonstrated end-to-end CAR-T production in early 2025, with the Premiere usine consumables facility still being brought up. Industrial orchestration across multiple sites is ahead of the platform rather than behind it.
Operational technology integration 42 → 86
The MTP direction is set and the SKALE project explicitly targets modular fleet operation, but the fleet layer and the instrument-level integration layer are still being specified. A working MTP service interface across vendors closes most of the remaining gap.
Data integrity and compliance 50 → 88
Research data capture is established, but decentralised commercial production requires a GxP-aligned electronic batch record on every unit and an ALCOA+ (Attributable, Legible, Contemporaneous, Original, Accurate, plus Complete, Consistent, Enduring and Available) audit trail that regulators can inspect from any site without an on-site presence.
Onboard analytical maturity 38 → 80
Fluorescence and brightfield imaging are planned and partially in place. The remaining gap is the onboard classification pipeline that turns raw images into a state and release-relevant signal in real time, including models trained on runs from multiple Lakhesys units.
Cloud-native scalability 35 → 82
Decentralised point-of-care deployment requires cloud-agnostic fleet management and analytics that can be operated under the data residency rules of different jurisdictions. Current systems sit closer to development laboratories than to a fleet-ready deployment.
Operator interface maturity 55 → 85
An intuitive operator interface capable of supporting a 30 percent reduction in manual errors is a stated target. The remaining gap is the live data tie-in between that interface and the run, which is where repetitive decision-making can be guided by the device rather than left to operator recall.

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 Astraveus, 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].