Bridging clinical cell therapy and global manufacturing

What we'd propose

• Digital Lab

  Computer vision monitoring of cultivation bioreactors

  Camera-based monitoring of the cultivation bioreactor, with foam detection, cell morphology analytics and contamination signatures running as a continuous second observer, so the cultivation cycle is characterised while the batch is still running.

  • Camera-based foam detection

    Foam levels read by vision, not by eye

    External cameras with OpenCV-based image processing detect foam levels in real time, distinguishing slow-rising from flash foam so the operator or the control system can act on the right kind of event.

  • Adaptive antifoam dosing

    Closed-loop dosing based on the visual signal

    A dosing controller running interval, vector or PWM modes adjusts antifoam addition against the visual foam signal, so dosing follows the current state of the bioreactor instead of a fixed schedule.

  • Cell morphology analytics

    AI-assisted read of cell condition

    Machine learning models trained on microscope and in-line images flag contamination signatures and cell health indicators before they are visible in offline samples, giving the cultivation team a lead on what offline assays will confirm.

  • The bioreactor is observed through every hour of the run, regardless of who is on shift.
  • Foam events trigger a dosing response in minutes, not at the next scheduled reading.
  • Contamination signals surface earlier in the run, shortening the time between detection and action.

  We've done this before

  • Bioreactor Control: Computer Vision for Non-Invasive Foam Management
  • Microalgae Cell Detection & Classification with an ML Model

  See how we do Digital Lab
• Digital Lab

  Real-time KPI engine and P&ID-based dashboards

  An automated KPI engine that reads directly from the process data stream and exposes the values in dashboards laid out like the physical bioreactor, with Golden Batch overlays so the current run is compared against the best historical profile.

  • P&ID-based dashboard layouts

    Dashboards that mirror the physical rig

    Dashboards built in Grafana to match the bioreactor's piping and instrumentation diagrams, reducing the cognitive distance between the screen and the equipment and shortening the path from observation to action.

  • Automated KPI calculation

    Bio-KPIs as they are measured

    Twelve or more biological KPIs, including Viable Cell Density, Viable Cell Volume, Growth Rate and CSPR, calculated inside the data platform directly from the instrument stream, so the numbers exist in the system rather than in a spreadsheet.

  • Golden Batch overlay

    Current run against the best historical run

    Overlay of the current cultivation profile against historical best runs, so deviations show up as a shape on the chart instead of as a number out of context.

  • Bio-KPIs arrive in the same view as the live process signals they describe.
  • Scientist self-service replaces the IT ticket that used to gate every dashboard change.
  • Deviations against the Golden Batch are visible as a curve, not after a meeting.

  We've done this before

  • Transforming Raw Bioprocess Data into Actionable Insights
  • Real-Time Process Intelligence & KPI Visualization

  See how we do Digital Lab
• Digital CDMO

  MTP-based equipment standard for distributed GMP sites

  A Module Type Package (MTP) and OPC UA equipment-integration baseline, with a NAMUR-compliant PLC library, that lets a new GMP site be onboarded against the same interface and validation package as the Gdańsk pilot.

  • MTP library implementation

    One PLC library for every equipment skid

    A PLC library compliant with VDI/VDE/NAMUR 2658, with object-oriented method modules, so a new skid is programmed against the same building blocks rather than from a blank project.

  • OPC UA backbone

    Open connectivity from instrument to platform

    OPC UA as the communication backbone between instruments, line supervision and the central data platform, with secure machine-to-machine communication and a documented information model.

  • Plug-and-produce validation pack

    Onboarding a new site in weeks, not quarters

    Pre-validated processing components and an integration guide that let any suitably certified GMP facility onboard the PEARL procedure against the same baseline, reducing the validation effort that has historically been per site.

  • The same equipment standard is reusable as new GMP sites join the network.
  • Vendor lock-in is reduced because the integration interface is the published standard, not a specific controller.
  • Onboarding a new site moves from a one-off project to a documented procedure.

  We've done this before

  • MTP Library for Modular Automation
  • Scaling Biotech Manufacturing via Universal IT/OT Architecture

  See how we do Digital CDMO
• Enterprise AI

  Zero Trust OT security for cloud-monitored GMP operations

  An identity-based access model for industrial automation, with active anomaly detection on the OT network and secure connectivity for air-gapped legacy instruments, so remote monitoring of distributed GMP sites happens inside an auditable security envelope.

  • Identity-based access control

    Verify at every access point

    Replace perimeter-only controls with identity verification at every access to the OT network, preventing lateral movement from a compromised credential into the manufacturing control plane.

  • Active anomaly detection on OT

    Distinguishing sensor faults from cyber events

    Continuous monitoring that separates a sensor glitch from an attack on the controller, so the operations team is not chasing false positives while a real intrusion goes unflagged.

  • Secure air-gap bridging

    Legacy instruments connected without exposure

    A secure connectivity framework that brings data from air-gapped legacy lab instruments into the central data lake without exposing proprietary methods or patient-derived material to the public network.

  • Remote monitoring of GMP sites happens inside a security envelope that has been designed for it.
  • Operational technology events are classified by cause, not just by alarm.
  • Legacy instruments contribute their data to the platform without joining the open network.

  We've done this before

  • Support Framework & Assisted Reality for Global Biotech Manufacturer

  See how we do Enterprise AI
• Agents

  AI agents for ATMP regulatory and quality documentation

  Narrow, reviewable agents that take the recurring part of EMA/FDA documentation work: drafting deviation and change-control summaries from source records, checking a document against its template before review, and finding every controlled document a standards change affects. A named person approves every output.

  • Drafting from source records

    First drafts of regulatory and quality documents

    Generate first drafts of deviation, change-control and periodic-review documents from the underlying system records, so the author edits and judges rather than assembles the document from scratch.

  • Template and completeness checking

    Gaps caught before review

    Check a submitted document against its EMA/FDA template and the site's own checklist, returning missing or inconsistent sections before the document enters the human review queue.

  • Change impact search across the document set

    Scope of a standards change, by search

    When a standard, method or specification changes, retrieve every controlled document 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 and traceable to their source records.
  • The scope of a standards change is established by search, not by memory.
  • Every agent output is reviewed and signed off by a named reviewer before it becomes a controlled document.

  We've done this before

  • Process Transfer Reporting System

  See how we do Agents