☰

Dipharma Francis S.r.l.

One data model across four cGMP sites

Industry
API Manufacturing and CDMO Services
Headquarters
Baranzate, Milan, Italy
Public information as of
March 2026

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

Strategic priorities

Dipharma Francis S.r.l. operates as a third-generation, family-owned API manufacturer and CDMO with four cGMP facilities across Italy and the United States: Baranzate (Milan) for headquarters and multipurpose manufacturing, Mereto di Tomba (Udine) for industrial-scale production, Caronno Pertusella (Varese) for generic APIs and advanced intermediates, and the Dipharma Inc. site in Kalamazoo, Michigan for North American R&D, kilolab and commercial supply. Commercial offices in China extend the network. The Mereto di Tomba site operates as a Seveso upper-tier facility, with the safety-report and regulatory regime that classification implies.

The organisation has said it is pursuing a 'vertical integration' strategy that links preclinical chemistry in Kalamazoo to commercial-scale API production in Italy, and the CEO has described an 'unprecedented increase in demand' for small-scale CDMO projects. The US footprint has expanded through phased investment: a dedicated cGMP kilolab and QC laboratory in phases 1 and 2 (2020 to 2022), then a second cGMP line in phase 3 (2023) with reactors up to 105 litres and expanded cryogenic and solid-isolation capability.

Digital maturity across the network is uneven. The Kalamazoo QC laboratories are designed to meet 21 CFR Part 11 compliance for electronic data integrity; legacy Italian sites continue to rely on manual data entry and paper-based workflows. Dipharma must satisfy the FDA (since 1970 at Baranzate, 1978 at Caronno, 1980 at Mereto), AIFA, PMDA, Korean FDA, ANVISA and Health Canada simultaneously, while defining its ESG strategy in line with the EU Corporate Sustainability Reporting Directive (CSRD) and working up from an EcoVadis Bronze medal in 2024.

Bringing the four sites onto one data and process model is the precondition for the integration story the leadership is telling. Where data moves between systems by hand, the response time to a customer request, an audit question or a Seveso safety check is set by the slowest transfer. The work ahead is to make process, quality and energy data readable across all four sites as one estate rather than four.

Challenges we see

  • Digital Integration

    Harmonising data integrity between Kalamazoo and the Italian sites

    Dipharma operates four cGMP sites with different digital baselines. The Kalamazoo QC laboratories are designed to meet 21 CFR Part 11 compliance for electronic data integrity, while Italian facilities continue to use manual data entry and paper-based workflows for many records.

    Where one site's records are electronic with audit trails and another site's are paper with manual transcription, the same batch looks like two different evidence chains. Reconciling them by hand lengthens every cross-site question.

  • Operations Manufacturing

    Monitoring hazardous reactions without putting sensors inside them

    The Mereto di Tomba site operates as a Seveso upper-tier facility handling high-energy materials such as nitroglycerine and running high-pressure cryogenic reactions, with 200+ personnel and 200 cubic-metre reactor capacity.

    Where internal probes are not practical inside a reactor holding an explosive charge, the population that is monitored is the operator looking through a sight glass. External sensors and edge models narrow that gap to the moments that matter.

  • Operations Manufacturing

    Transferring process parameters between kilolab and industrial scale

    The Kalamazoo site runs kilolab reactors up to 105 litres and the Mereto di Tomba site runs industrial reactors at 200 cubic metres; the CEO has publicly cited an unprecedented increase in demand for small-scale projects.

    When a process moves from a 105-litre reactor to a 200-cubic-metre one, the parameter mapping is mostly held in the heads of the engineers who have done it before. Making it visible to the next person on the project is what turns experience into a transferable asset.

  • Compliance Regulatory

    Preparing for multilateral audits on demand

    Dipharma must satisfy the FDA, AIFA, PMDA, Korean FDA, ANVISA and Health Canada across four sites while preserving a 50-year record of positive inspections. A first Pre-Approval Inspection at the US site was completed in 2023.

    When a regulator asks for evidence on a given lot, the answer depends on how quickly each site's records can be assembled into one query. The audit response time is a property of the data model more than the regulatory team.

  • Digital Integration

    Bringing site-level OT under cybersecurity and safety standards

    Global ICT is managed centrally, while industrial control systems at the site level retain site-specific security profiles. Industry 4.0 connectivity is increasing and the regulatory environment now includes NIS2 (the EU directive on cybersecurity of network and information systems) and IEC 62443 (the international standard for industrial cybersecurity).

    As more plant data leaves the controller, the security boundary moves with it. Setting zones, conduits and remote-access rules before new connections are added is what keeps the boundary explicit instead of improvised.

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. Building one data model across all four facilities

    Kalamazoo records are electronic with audit trails and Italian records are often paper-based, so a single batch lives in different shapes across sites. The result is that cross-site queries are assembled by hand from different source systems.

    An ontology-based industrial data platform that defines assay, batch, equipment, lot and site once, then loads records from each facility against that model, lets scientists overlay live process trends against historical 'golden batches' across sites without a per-query integration project.

    • Dipharma research, organisational cartography and IT/OT fragmentation notes, 2024
    • Dipharma sustainability and Kalamazoo 21 CFR Part 11 references

  2. Monitoring hazardous reactors from the outside

    Direct internal instrumentation of reactors containing high-energy materials is impractical because of fouling, heat and the risk of ignition. Conventional inspection at Mereto di Tomba is therefore periodic and human.

    External cameras and edge-deployed models trained on vibration, sound and thermal signatures can flag a deviation in the seconds before it becomes a safety event, with a digital twin of the reactor giving the operator a view that mirrors the physical P&ID (piping and instrumentation diagram).

    • Dipharma Seveso upper-tier status and hazardous chemistry references
    • Mereto di Tomba site profile, 251,200 sq ft facility

  3. Capturing scale-up parameter transfers as data

    Moving a process from the 105-litre Kalamazoo kilolab reactors to the 200-cubic-metre Mereto reactors depends on the judgement of engineers who have done it before. The mapping is rarely stored as data the next team can read.

    Recording scale-up runs as structured process records, including the parameter choices and outcomes, builds a library that the next team can search and reason over instead of asking the same colleagues the same questions.

    • Dipharma vertical integration strategy and Kalamazoo phase 3 expansion
    • CEO statement on demand for small-scale projects

  4. Standardising equipment interfaces with MTP and OPC UA

    Equipment changeover between different API synthesis contracts can take weeks at multipurpose facilities, limiting the agility that Dipharma describes as a core competitive advantage. Each reactor speaks its own vendor protocol.

    Implementing the Module Type Package (MTP) standard with OPC UA (Open Platform Communications Unified Architecture) as the communication backbone lets each reactor behave as a documented service that the line can compose in hours rather than weeks.

    • Dipharma 'flexibility and agility' positioning, Company Profile 2024
    • Dipharma IT/OT fragmentation analysis

  5. Automating CSRD and EcoVadis reporting from the line

    EcoVadis Bronze in 2024 is a starting point for Dipharma's sustainability programme, and EU CSRD compliance will require evidence that is currently assembled by hand from fragmented environmental data sources across four sites.

    An energy and environmental data pipeline that reads consumption directly from cogeneration, trigeneration and solar systems turns CSRD reporting from a periodic exercise into a continuous readout, and supports the move toward Silver and Gold EcoVadis ratings.

    • Dipharma sustainability statement, 2024
    • EcoVadis Bronze medal, 2024
    • EU Corporate Sustainability Reporting Directive (CSRD)

What we'd propose

  • Enterprise AI

    One industrial data platform across all four sites

    An ontology-based data platform that defines batch, equipment, lot, assay and site once, then loads records from Baranzate, Mereto di Tomba, Caronno Pertusella and Kalamazoo against the same model, so cross-site queries and 'golden batch' comparisons happen on shared data.

    • Cross-site data ontology

      One shared semantic model

      Build a semantic layer that maps production parameters from each site's level-2 and level-3 systems onto a common set of entities, so a query written once returns comparable answers from Baranzate, Mereto, Caronno and Kalamazoo instead of four dialects of the same table.

    • Pipelines from each site's source systems

      Loading every facility

      Build ingestion for Kalamazoo's electronic batch records, the Italian sites' LIMS (Laboratory Information Management System) and ELN (Electronic Laboratory Notebook) output, and equipment historians, with schema validation at the boundary so a malformed record fails loudly instead of silently.

    • Cross-site golden batch comparison

      Live against historical

      Overlay current run trends against historical best runs from any facility on a single dashboard, so the next engineer approaching a scale-up question starts from prior evidence rather than prior recollection.

    • Cross-site questions stop being a per-request integration project.
    • The vertical integration story from Kalamazoo to Mereto is backed by data that follows the molecule.
    • Future sites and acquisitions attach to the model rather than triggering another migration.

    We've done this before

    See how we do Enterprise AI
  • Digital CDMO

    External monitoring for hazardous reactors

    An external sensing and edge-AI system that watches Mereto di Tomba's hazardous reactors from outside the vessel, predicting dangerous deviations from vibration, sound and thermal patterns, with a digital twin view for the operator.

    • Non-invasive sensor package

      Cameras and edge devices on the outside

      Mount cameras and accelerometers around the reactor exterior so the process is observed without contact with the chemistry, eliminating the fouling and ignition risk an internal probe would carry.

    • Edge AI for early deviation detection

      Seconds before an event

      Run compact machine-learning models on the edge device that flag vibration, acoustic and thermal signatures drifting from a learned normal envelope, so the operator gets a signal in seconds rather than a finding after the batch.

    • Digital twin of the reactor for the operator

      A view that matches the P&ID

      Mirror the live readings onto a digital twin of the reactor laid out like the physical P&ID (piping and instrumentation diagram), so the operator's mental model of the plant carries over to the screen without translation.

    • Seveso-zone monitoring becomes continuous instead of periodic.
    • Operators see the same view of the reactor the safety system sees.
    • A deviation is detected in time for the operator, not after the batch.

    We've done this before

    See how we do Digital CDMO
  • Digital Lab

    Digital lab and QC transformation across sites

    A programme that brings R&D and QC laboratories at all four sites onto shared electronic data capture with self-service analytics, replacing paper-based and Excel-based workflows while keeping ALCOA+ (the data-integrity expectations of 'attributable, legible, contemporaneous, original and accurate, plus complete') and 21 CFR Part 11 compliance.

    • Electronic lab notebook and LIMS rollout

      Paper replaced by audit-trailed records

      Deploy an ELN and LIMS in Kalamazoo and the Italian QC labs so every result is captured with method version, instrument identity and timestamp, and so paper logbooks stop being the legal record.

    • Self-service KPI dashboards

      Scientists build their own charts

      Stand up a Grafana-based analytics layer that lets scientists build their own charts against the lab data without filing an IT request, reducing the manual post-experiment analysis that drives most 'Excel islands'.

    • 21 CFR Part 11 evidence chain

      Records that stand up in inspection

      Implement electronic signature, versioning and audit-trail handling to 21 CFR Part 11 across all four sites so a regulator inspecting any of them sees the same evidence chain Kalamazoo already runs.

    • Scientist time moves from transcription to interpretation.
    • Scale-up runs are recorded as data the next team can read, not memory the next team has to ask for.
    • Italian QC labs operate to the same electronic standard as Kalamazoo.

    We've done this before

    See how we do Digital Lab
  • Digital CDMO

    MTP-based modular production framework

    A Module Type Package (MTP) library and the OPC UA backbone that connects it, allowing glass-lined and stainless steel reactors to be brought onto new campaigns as documented services with predictable interfaces.

    • MTP-compliant module library

      Standardised control modules

      Build the MTP-compliant module descriptions for the reactors and peripheral units in use at Baranzate, Mereto and Caronno, so each piece of equipment exposes its control face as a documented service rather than a vendor-specific block.

    • OPC UA as the line backbone

      One protocol across vendors

      Stand up OPC UA (Open Platform Communications Unified Architecture) as the communication backbone across the multipurpose lines, so the SCADA (Supervisory Control and Data Acquisition) layer and any new equipment speak the same language without per-vendor integration work.

    • Plug-and-produce line reconfiguration

      Hours rather than weeks

      Combine the MTP library and OPC UA backbone into a workflow where a new campaign's reactor set is composed on screen, validated, and handed to operations in hours instead of being engineered over weeks.

    • Campaign changeover time compresses from weeks to hours.
    • New equipment is integrated by configuration rather than by project.
    • The same module language is reusable at every multipurpose site.

    We've done this before

    See how we do Digital CDMO
  • Agents

    AI agents for regulatory and sustainability documentation

    Narrow, reviewable agents that take the repetitive part of regulatory and sustainability documentation across the four sites: drafting responses from source records, checking a submission against the regulator's template, and tracking the documents a standards change touches. A named person approves every output.

    • Drafting from source records

      First drafts from system data

      Generate first drafts of FDA, AIFA, PMDA and ANVISA responses and periodic review reports from the underlying batch, deviation and change-control records, so the regulatory affairs team edits and judges rather than assembles.

    • Template and completeness checking

      Gaps found before review

      Check a draft against the template and the site's own checklist before it enters the human review queue, returning missing or inconsistent sections so reviewers see a complete document instead of an unfinished one.

    • CSRD and EcoVadis evidence assembly

      Sustainability data pre-staged

      Pull energy, water and waste readings from the line and pre-stage the evidence packs for CSRD reporting and the next EcoVadis submission, so the sustainability team signs off a prepared dataset rather than building one from spreadsheets.

    • Regulatory response time compresses because drafts arrive complete.
    • CSRD evidence is a continuous readout, not a year-end exercise.
    • Every output is traceable to the source records and signed off by a named reviewer.

    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 Dipharma Francis S.r.l.'s own published ambition implies — not a perfect score.

Source: A4BEE analysis of public sources
Data integration 40 → 85
Kalamazoo operates to 21 CFR Part 11 electronic data integrity while Italian sites continue to rely on manual data entry and paper-based workflows. The cross-site gap sets the baseline.
Process automation 45 → 80
Italian sites run air-gapped or legacy control systems with periodic human inspection, while the most hazardous reactors at Mereto are still observed through sight glasses rather than continuous external sensing.
Analytics and intelligence 35 → 75
Cross-site 'golden batch' comparison and live overlay of process trends depend on data that is not yet shared between facilities; analytical work largely happens offline in spreadsheets.
IT and OT convergence 30 → 70
Corporate ICT is managed globally but site-level OT retains vendor-specific configurations; a unified connectivity backbone is still ahead of the operating model rather than behind it.
Cybersecurity 40 → 80
NIS2 and IEC 62443 compliance for the OT layer is a current requirement rather than a future target, and the security boundary is moving as more plant data leaves the controller.
Sustainability tracking 35 → 70
EcoVadis Bronze in 2024 is the baseline; CSRD reporting will require environmental data that is currently assembled by hand from disparate sources across four sites.

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 March 2026. It reflects A4BEE's own interpretation and opinion, is not affiliated with, endorsed by, or verified with Dipharma Francis S.r.l., 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].