Apoha
Last reviewed
Jun 3, 2026
Sources
6 citations
Review status
Source-backed
Revision
v1 · 1,387 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 3, 2026
Sources
6 citations
Review status
Source-backed
Revision
v1 · 1,387 words
Add missing citations, update stale details, or suggest a clearer explanation.
Apoha is a deep-technology startup, based in London and San Francisco, that is building AI models for designing and discovering new substances by measuring how materials physically behave rather than relying only on their chemical composition or structure. The company emerged from stealth on 3 June 2026 with a $36 million Series A round led by the European venture firm Singular, and it frames its core method as "liquid intelligence": a way of capturing the "wave forms" a tiny sample of material throws off when it is suspended in a liquid and pushed by a controlled sequence of physical stresses.[1][2] Apoha argues that this kind of measurement, which it calls a third fundamental data class for molecular science alongside sequence and structure, does not yet exist at scale and cannot be scraped from the internet or copied from existing assays. That positioning is the company's own, and the underlying science remains early.[2][4]
Apoha sits at the intersection of AI for materials discovery, laboratory instrumentation, and biophysics. Most computational efforts to design proteins, drugs, foods, or other materials lean on data about what a molecule is made of (its sequence) or what it looks like (its three-dimensional structure). Apoha's bet is that a third type of data, how a material actually behaves under stress, carries information the other two miss, including properties such as aggregation, binding, folding, degradation, and even sensory qualities like smell and taste.[1][4]
To collect that data, the company built laboratory hardware that takes a sample small enough to sit on the head of a pin, suspends it in a liquid, and applies a controlled series of tiny perturbations at the liquid interface. The instrument records the wave patterns that ripple back through the liquid in response. According to Apoha, a single run yields more than 1,000 distinct numerical descriptors of how the material behaves, and it takes minutes rather than the days or weeks that conventional, one-property-at-a-time lab tests require.[1][2] The company's first product, VIBE (which it expands as "Variations in Interfacial Behaviour under Excitation"), produces what it describes as a high-dimensional behavioural signature: a representation of how a sample behaves rather than what it is made of.[2][4]
The longer-term goal is to feed enough of this wave data into AI models so they can run in reverse. Instead of only measuring an existing material, the models would suggest how to modify or create one to hit a target set of properties. Apoha calls this combination of new measurements and learned models "liquid intelligence," and its investors and marketing also use the branding "Liquid Brain" for the underlying platform.[1][3]
The technical foundation traces back to work that chief executive Shamit Shrivastava began around 2008 on interfacial physics and thermodynamics.[2] Shrivastava holds a PhD from Boston University and did postdoctoral research in the Department of Engineering Science at the University of Oxford, where he studied how nonlinear acoustic pulses move along lipid interfaces.[1] In peer-reviewed work published in the Journal of the Royal Society Interface in 2018, he and collaborator Matthias Schneider showed that nonlinear sound waves traveling along a lipid monolayer near a phase transition could collide and annihilate one another, reproducing several hallmarks of nerve impulses, including threshold behaviour, all-or-none firing, and a fixed propagation velocity, purely from thermodynamic principles applied to an interface.[5]
That research connects to a broader, and still debated, line of inquiry sometimes called the soliton theory of the nervous impulse, which proposes that nerve signals are mechanical waves in the cell membrane rather than purely electrical events. Apoha's commercial insight is narrower: if a wave traveling through matter near a phase boundary is shaped by the thermodynamic state of the material it passes through, then recording that wave gives you a rich, indirect readout of the material's properties. The company describes the interface as "an amplifier of state rather than a passive container."[4] More recent preprints posted to bioRxiv in 2025 applied this interfacial-wave approach to profiling therapeutic antibodies, reporting that the resulting signal captured a composite of hydrophobicity, self-interaction, and thermal stability rather than any single conventional measurement.[6]
Apoha was co-founded in 2021 by Shrivastava, who serves as chief executive, and Anshika Srivastava, the chief operating officer and a former executive director at Goldman Sachs.[2] Shrivastava brings the underlying physics and instrumentation, having spent more than a decade on the interfacial-wave problem in academia before commercialising it; Srivastava brings the operating and finance background. The company says it holds a large patent portfolio, reported at more than 60 filings spanning hardware, software, data, and the AI models themselves, although patent counts are self-reported and many filings may still be pending.[2]
By the time it left stealth, Apoha had grown to roughly 25 employees split between its London and San Francisco offices and said it had completed around 40 customer projects.[1] It announced the funding from the Frontier Technologies Stage at the SXSW London event on 3 June 2026.[2]
The $36 million Series A was led by Singular, a European venture capital firm, with participation from Tim Draper's Draper Associates. Existing seed investors Redalpine, Seedcamp, Wilbe, and Nucleus continued their backing, and the company also holds a grant from Innovate UK, the United Kingdom's national innovation agency.[1][2] Apoha did not disclose its valuation. The round is being used to expand the data-collection hardware, grow the AI modelling team, and push further into commercial work across pharmaceuticals, food, and materials.[1]
Although Apoha is brand new as a public company, it points to several named collaborations to argue that its measurements have practical value. The most concrete is a multi-year commercial partnership with the pharmaceutical company Boehringer Ingelheim. In joint research, Apoha reported that its VIBE platform identified high-risk therapeutic antibody candidates with greater than 90% precision using as little as 8 to 10 micrograms of material, a fraction of what standard developability assays consume.[1][2] These results were described in 2025 bioRxiv preprints and have not yet completed peer review, so they should be read as promising early data rather than settled findings.[6]
Apoha has also disclosed work with the German biotech firm Ethris on predicting how lipid nanoparticles carrying mRNA behave in animals, a collaboration with the plant-based food company THIS on finding a protein replacement, and a relationship with Somru BioSciences. The company says additional customers include several Fortune 500 firms across pharmaceuticals, food, and materials, though those are not named.[1][2]
Apoha enters a crowded and fast-moving field. Machine learning for materials discovery has drawn heavy investment since DeepMind's GNoME work predicted hundreds of thousands of candidate crystal structures, and a wave of AI startups now apply generative models and autonomous labs to drug design, battery chemistry, and synthetic biology. Most of these efforts share a common bottleneck: the quality and quantity of experimental data available to train on. Sequence and structure databases are large, but data on how materials actually behave under real conditions is comparatively scarce and expensive to generate.
What distinguishes Apoha's pitch is the claim that it has built a cheap, fast, label-free way to generate that behavioural data at scale, and that the data is genuinely new rather than a remix of what already exists. Whether the wave-form descriptors generalise across very different classes of materials, and whether the physics-based interpretation holds up under independent scrutiny, are open questions that the next few years of customer work and peer review will test. For now, the funding and the named partnerships suggest serious investors and at least one large pharmaceutical company are willing to bet that the approach is worth pursuing.