You open a browser tab to a cloud dashboard. The interface is clean, almost sterile, with charts and sliders arranged like the control panel for a very expensive, very delicate machine. You are not piloting a spaceship. You are, according to the platform, modeling the downstream purification of a monoclonal antibody, tweaking pH and conductivity to maximize yield. The software suggests a parameter adjustment, its confidence interval shaded in light blue. It feels less like using an app and more like consulting a deeply knowledgeable, slightly pedantic colleague who speaks exclusively in differential equations. This is the user experience of a process engineer at one of the world’s largest pharmaceutical companies, and it is being built by DataHow, a nine-year-old spinout from ETH Zurich [DataHow, May 2024].
DataHow’s core product, DataHowLab, is not another generic AI copilot. It is a specialized digital workbench for the scientists and engineers who develop and scale the biological processes that manufacture modern drugs. Its wedge is a technical philosophy called hybrid modeling, which deliberately marries two often-opposed worlds: first-principles, mechanistic understanding of biochemistry, and purely data-driven machine learning [DataHow, September 2026]. The platform ingests experimental data from bioreactors and chromatographs, but it constrains its AI models with the known laws of physics and chemistry that govern those systems. The result, the company argues, is a predictive tool that requires fewer costly lab experiments, accelerates time to market, and provides a defensible rationale for every recommendation,a non-negotiable in the heavily regulated world of biopharma.
The digital workbench for making drugs
The problem DataHow is solving is one of immense complexity and cost. Developing a manufacturing process for a biologic drug,like the mRNA in vaccines or antibody-based therapies for cancer,is a painstaking, iterative science. Teams run countless small-scale experiments to find the optimal conditions for cell growth, protein expression, and purification. A single misstep in scaling this process from lab bench to factory vat can cost millions and delay life-saving treatments. For decades, this work has been guided by experience, intuition, and statistical design-of-experiment software. DataHow’s bet is that a hybrid-AI layer can become the new central nervous system for this entire development and production lifecycle, from initial screening to real-time control of live manufacturing [DataHow, Unknown].
The platform’s stated capabilities read like a roadmap for digitizing the physical lab:
- Process analytics and modeling. Turning raw sensor data into predictive models of how a process will behave under new conditions.
- Transfer learning. Applying insights from one molecule or process to accelerate development of another, a key efficiency for drugmakers with large pipelines.
- Real-time connectivity and control. Linking digital models directly to manufacturing equipment to enable adaptive, model-driven process control [DataHow, September 2026].
This last point is the focus of DataHow’s most recent bridge financing, secured in September 2026. The extension of its Series A, again led by industrial-focused venture firm Momenta, is earmarked for expanding into downstream processing and strengthening these real-time control features [DataHow, September 2026]. It’s a signal that the company is moving beyond the development lab and into the heart of the factory floor.
Traction in a fortress market
Selling enterprise software to large pharmaceutical companies is notoriously difficult. Sales cycles are long, compliance hurdles are high, and incumbents are entrenched. DataHow’s most compelling traction metric, therefore, is not a revenue figure but a customer roster. The company reports that its platform is now used by 12 of the top 20 global pharmaceutical companies [TetraScience, November 2026]. While the specific names are not disclosed, this claim, made in the context of a partnership announcement with data platform TetraScience, suggests a level of adoption that bypasses the typical startup pilot-project purgatory. The platform is reportedly deployed at scale within many large manufacturers [CB Insights, Unknown].
This traction has been built by a team rooted in the very academia it seeks to augment. DataHow is a spin-off from the Institute for Chemical and Bioengineering at ETH Zurich, founded in 2017 by a group of researchers including CEO Dr. Alessandro Butté, CTO Fabian Feidl, COO Michael Sokolov, and President Massimo Morbidelli, a professor at the university [DataHow, May 2024][Pharma Almanac, retrieved 2026]. The founding narrative is classic deep tech: years of research crystallizing into a commercial product, backed by investors who specialize in industrial transformation, like Momenta and Rockwell Automation [Greater Zurich Area, May 2024].
| Founder | Role | Background Anchor |
|---|---|---|
| Dr. Alessandro Butté | Co-Founder & CEO | ETH Zurich spinout lead; quoted as CEO in financing announcements [DataHow, September 2026]. |
| Fabian Feidl | Co-Founder & CTO | Co-founded DataHow AG in 2017; leads technology [Pharma Almanac, retrieved 2026]. |
| Michael Sokolov | Co-Founder & COO | Co-founder status reported in financing coverage [The Venture Codex, retrieved 2026]. |
| Massimo Morbidelli | Co-Founder & President | Professor at ETH Zurich; serves as President [Bioprocessing Summit Europe, retrieved 2026]. |
The integration imperative
The primary competitive pressure on a platform like DataHowLab is not a direct, feature-for-feature clone. It is the inertia of legacy workflows and the sprawling, fragmented nature of biopharma IT. A process engineer’s digital life is already a patchwork of electronic lab notebooks (ELNs), laboratory information management systems (LIMS), historian data stores, and proprietary equipment software. A new AI platform, no matter how clever, is just another silo if it cannot connect. DataHow’s answer has been a deliberate partnership strategy, embedding its models into the ecosystems where scientists already work.
Its partnership with TetraScience, announced in late 2026, is a prime example. TetraScience operates a cloud platform for scientific data, serving as a central hub for experimental data across major pharma R&D labs. The integration allows DataHowLab to pull validated data directly from that hub, making the AI modeling step a smooth part of an existing data flow [TetraScience, November 2026]. Similarly, a strategic collaboration with lab equipment giant Eppendorf aims to streamline data management from Eppendorf’s bioreactors directly into DataHow’s models [DataHow, December 2024]. These are not mere press release partnerships; they are distribution and distribution channels, reducing friction at the point of ingestion.
The next twelve months
The bridge financing from Momenta sets a clear agenda. The next phase for DataHow is about moving from a powerful development tool to an indispensable production system. The focus on real-time connectivity and downstream processing points to a future where DataHowLab’s models are not just used to design a process, but to autonomously control and optimize it as it runs in a multi-million-dollar Good Manufacturing Practice (GMP) facility. Success here would transform the company’s value proposition from cost-saving and speed to direct impact on product quality and supply chain reliability.
The company is also hiring for roles like Lead Machine Learning Engineer, indicating a build-out of the core technical team to support this ambition [DataHow, retrieved 2026]. The key milestone to watch will be public case studies or further partnership announcements that specifically cite live manufacturing control deployments, not just development-stage use cases.
For an industry built on the meticulous, physical manipulation of living cells, the idea of a digital twin feels both inevitable and faintly heretical. DataHow’s bet is that the heresy is now a necessity,that the complexity and cost of making biologic drugs have outstripped the capacity of human intuition alone. The cultural question the platform is implicitly answering is not whether AI belongs in the lab, but what kind of AI. In a field where a model’s recommendation must be explainable to a regulator, a pure black-box neural network is a non-starter. DataHow’s hybrid approach offers a compromise: the predictive power of machine learning, wearing the respectable, equation-filled suit of mechanistic science. It is building the AI that a process engineer can actually trust.
Sources
- [DataHow, May 2024] DataHow secures Series A investment for its AI bioprocessing solutions | https://datahow.ch/news/datahow-secures-series-a-investment-for-its-ai-bioprocessing-solutions/
- [DataHow, September 2026] DataHow secures bridge financing to expand its digital bioprocess development solution | https://datahow.ch/news/datahow-secures-bridge-financing-to-expand-its-digital-bioprocess-development-solution
- [TetraScience, November 2026] DataHow joins the Tetra Partner Network | https://www.tetrascience.com/news/datahow-joins-the-tetra-partner-network
- [Greater Zurich Area, May 2024] DataHow successfully closes financing round | https://www.greaterzuricharea.com/en/news/datahow-successfully-closes-financing-round
- [DataHow, Unknown] Home - DataHow | https://datahow.ch/
- [CB Insights, Unknown] DataHow company profile | https://www.cbinsights.com/company/datahow
- [DataHow, December 2024] Eppendorf SE and DataHow AG announce strategic collaboration to advance bioprocess data management | https://datahow.ch/news/eppendorf-se-and-datahow-ag-announce-strategic-collaboration-to-advance-bioprocess-data-management/
- [Pharma Almanac, retrieved 2026] Fabian Feidl profile | https://www.pharmaalmanac.com/
- [The Venture Codex, retrieved 2026] Michael Sokolov profile | https://theventurecodex.com/
- [Bioprocessing Summit Europe, retrieved 2026] Massimo Morbidelli profile | https://bioprocessintl.com/
- [DataHow, retrieved 2026] Careers page | https://datahow.ch/about/careers/