AI FOR SCIENCE · COMPANIES
Companies
Every organization AI4S.fyi keeps a page for: the research, interviews, people and rounds the record holds on it, each read at source. A count is what the record holds, not a score.
193 organizations
Frontier labs and platform companies 11
Each carries AI4S.fyi's own analysis of its direction.
AnthropicBets that the scarce thing in AI-assisted science is not candidates but verification, and builds the checkers, interfaces and access that make a claim inspectable (since September 2026, in biology, the bench as well).45 timeline · 1 interview · 38 people · Latest 
Google DeepMindTen public directions, from a 9-billion-variant prediction database to closed-loop synthesis on a real CVD reactor. The London materials lab is being built; drug design sits in Isomorphic Labs.76 timeline · 2 interviews · 50 people · Latest 
Microsoft ResearchUses learned simulators and generative models to shift the speed–accuracy frontier, then enters R&D through Azure, open code and industry partnerships.10 timeline · 20 people · Latest 
OpenAIPuts general agents inside the tools researchers already open (code, instruments, the paper itself) and measures progress in research throughput rather than domain models.27 timeline · 2 interviews · 13 people · Latest 
Isomorphic LabsExtends AlphaFold-style structural learning into a closed drug-design engine for pharma partnerships and proprietary small-molecule assets.9 timeline · 11 people · 3 rounds · Latest 
Lila SciencesTurns multi-domain automated laboratories into training environments where physical experiments produce verifiable scientific tokens.5 timeline · 4 interviews · 31 people · 4 rounds · Latest 
Radical AIUses an AI Scientist + Self-Driving Lab to move materials R&D from predicted candidates toward physical experiments, manufacturing and final applications.3 timeline · 1 interview · 4 people · 2 rounds · Latest 
FutureHouseMakes literature search, evidence synthesis and research planning measurable scientific agents, then commercializes them through Edison.6 timeline · 3 people · Latest 
Anew LabsByteDance's AI pharmaceutical team, spun out in 2026. Known publicly for Protenix, an open-source AlphaFold3 reproduction; internally it wires structure prediction, generative design, affinity and R&D reasoning into one closed-loop engine, with IL-17 and IL-4R in the pipeline.4 timeline · 1 person · 1 round · Latest 
RecursionIndustrializes cellular phenomics to create proprietary data, maps disease–target–compound relationships and monetizes through milestones and a clinical pipeline.3 timeline · 25 people · Latest 
DP TechnologyBuilds on molecular simulation methods such as DeePMD to connect research cloud, drug and materials software, scientific agents and experimental facilities.2 timeline · 4 people · 7 rounds · Latest
AI4S startups 118
Companies on the funding board, most recently active first.
- ALLOXMeasures the effects of thousands of protein mutations at once to map allosteric and other regulatory sites, and builds AI models on that data to find drug-binding spots on hard targets such as protein-protein interactions and transcription factors, for its own small molecules.1 timeline · 6 people · 1 round · Latest
- BioGeometryBuilds generative models for biology: GeoFlow, an all-atom model that predicts the structures of proteins and their complexes and designs binders, and GeoCell, a virtual-cell model of how cells' gene expression responds to genetic and drug perturbations. An automated lab expresses and tests the designs and feeds the data back. It designs antibodies, vaccines, peptides and enzymes for partners and its own pipeline, including synthetic-biology products.1 timeline · 6 people · 3 rounds · Latest
- Atlas BioRuns in silico clinical trials, starting in immunology: a foundation model trained on about 2 million RNA readouts from human tissue predicts each virtual patient's response to several treatments and diseases, to help drug developers choose what to take into Phase II.1 timeline · 3 people · 2 rounds · Latest
BigHat BiosciencesDesigns antibodies with machine learning tied to a fast automated wet lab (Milliner) that makes and tests designs; its lead drug BHB810 is in the clinic.4 timeline · 9 people · 3 rounds · Latest - BiographicaUses genomic foundation models, pretrained on public sequencing data and fine-tuned on data from its own lab, to predict which crop genes control a trait and how to edit them, and supplies those gene targets to seed and breeding companies across crops.1 timeline · 3 people · 1 round · Latest
CuspAIBuilds AI models that search for new materials with properties a customer asks for, first for carbon capture and now mostly for semiconductor materials.4 timeline · 1 interview · 19 people · 3 rounds · Latest 
EnvedaUses mass spectrometry and machine learning (PRISM) to identify drug candidates among natural compounds from plants and other organisms, then develops them itself.5 timeline · 13 people · 9 rounds · Latest 
Generate:BiomedicinesA Flagship company using generative AI to design protein drugs such as antibodies, peptides and enzymes, and developing them itself.6 timeline · 3 people · 3 rounds · Latest - OmabitSoftware that analyses a drug developer's own preclinical, clinical and genomic data alongside published research: a scientist chooses the question in plain language and it runs the analysis and writes it up, starting with clinical development, biomarker and translational teams.1 timeline · 2 people · 1 round · Latest
adsilicoA University of Leeds spinout whose Virtual Chimera platform builds virtual hearts and blood vessels from anonymised clinical scans with generative AI, then runs physics simulations of how a heart valve, stent or other implant behaves in each, across a population varied by age, sex, ethnicity, body size and disease.1 timeline · 5 people · 1 round · Latest 
Basecamp ResearchCollects genetic data from nature through partnerships in over 30 countries, trains its EDEN models on it, and now uses them to design therapies, starting with in vivo cell therapy.3 timeline · 8 people · 3 rounds · Latest - Pathos AILicenses clinical-stage drugs from other companies and develops them itself, using an agent platform (Foundry, earlier PathOS) trained on patient records to choose assets, pick the patients most likely to respond and adjust trials as they run. Its pipeline is in cancer: prostate cancer, multiple myeloma, lung and breast cancer and TROP2/HER3-expressing solid tumours; in October 2026 it began work in Alzheimer's disease with Bristol Myers Squibb and Tempus.1 timeline · 5 people · 3 rounds · Latest
Xaira TherapeuticsBuilds AI models for protein and antibody design and for cell biology, trained partly on its own large perturbation datasets, and runs its own drug programs.5 timeline · 13 people · 1 round · Latest - Zhichong PharmaDevelops new drugs for pets with generative AI, using its own small-molecule and large-molecule design engines (ChemoFetch and OmniFetch), and argues that testing in the target animal closes the loop from design to evidence faster than human drug development can; its programmes are in anti-infectives, chronic disease and ageing.1 timeline · 2 people · 2 rounds · Latest
ScineticsBuilding a science foundation model that reasons over DNA, RNA and protein directly, tied to agents that run robot-arm lab work.3 timeline · 3 people · 1 round · Latest 
Warp SciencesA US chemistry contract research organisation that makes the molecules drug-discovery teams order, starting with variations of a lead compound for lead optimisation. Chemists guide the work while the company automates more of it with AI and robotics and trains models on the data its experiments produce. It says it has no drug pipeline of its own.1 timeline · 4 people · 1 round · Latest 
AI ProteinsDesigns miniproteins from scratch with AI, then makes and tests them in a high-throughput lab, as drugs that are smaller than antibodies; its own pipeline starts in oncology.3 timeline · 4 people · 1 round · Latest 
MirendilTrains models meant to do AI research themselves and builds a platform around them, so labs in drug discovery, chemistry or biology can do frontier AI work without becoming AI labs.2 timeline · 3 people · 2 rounds · Latest 
ZelixirUses AI protein structure prediction and enzyme design to build synthetic-biology production routes; first product is an enzyme-made natural flavour made at a Changzhou plant.4 timeline · 3 people · 3 rounds · Latest - Electric PlantMakes electrodes and low-cost sensors that record a living plant's electrical signals from its stem or trunk, and Bombadil, a model trained on those signals across about a hundred species that estimates leaf stress measures such as photosynthetic efficiency and water stress, for farming, forestry and plant care.2 timeline · 2 people · 1 round · Latest
Metis ReasoningEarly-stage lab building AI meant to reason and act reliably beyond its training data, naming embodied AI, science and engineering as target fields.2 timeline · 4 people · 1 round · Latest 
Radical NumericsAn AI lab building general biological intelligence: foundation models trained on DNA and other biological data, released to partners as APIs and tuned models.6 timeline · 14 people · 1 round · Latest 
MATARIABuilds a multiscale materials model, from atoms to finished parts, that proposes candidates, narrows them with automated DFT and finite-element tools, and is tested by instrumented experiments; composites come first, then semiconductor and infrared materials.1 timeline · 4 people · 1 round · Latest 
METiS TechBioUses AI plus high-throughput lab work to design lipid nanoparticles, mRNA sequences and drug formulations, licensing the platform and its own programs to drugmakers.4 timeline · 10 people · 7 rounds · Latest - HolyvoltTrains battery-materials models on 18 years of cell-cycling data from the high-throughput lab it acquired, Wildcat Discovery Technologies, and lets the models choose what the lab makes and cycles next. Its first material is DRX, a cathode family with neither nickel nor cobalt.1 timeline · 2 people · 1 round · Latest
Terray TherapeuticsMeasures billions of small molecule-target interactions on dense microarrays and trains AI models (tNova, COATI) on that data to design small-molecule drugs.3 timeline · 7 people · 3 rounds · Latest 
XtalPiCombines physics-based computation, AI models and robotic labs to run drug discovery and chemistry projects for partners, and sells lab robots and software.8 timeline · 5 people · 4 rounds · Latest - Gaowei YiyaoMines protein-interaction and other high-dimensional biology data with its own models to find drug targets others treat as undruggable, then runs its own small-molecule and antibody programmes. Its GROOVY platform found the target behind GW-01, a Notch-pathway inhibitor now in Phase I; ACUMEN maps pairs of neighbouring cell-surface proteins for bispecifics and NEXUS looks for tumour neoantigens.1 timeline · 1 person · 1 round · Latest
Iambic TherapeuticsDesigns small-molecule cancer drugs with AI structure-prediction models (NeuralPLexer) linked to an automated chemistry and biology lab.6 timeline · 13 people · 5 rounds · Latest 
Leash BioCollects its own protein-chemical binding data with DNA-encoded libraries, and trains models and agent environments on it to solve medicinal chemistry.4 timeline · 1 interview · 6 people · 1 round · Latest - ScimplifyDevelops manufacturing processes for specialty chemicals and critical materials in its own labs in Bengaluru and Hyderabad, then transfers them to a network of more than 500 partner factories. Its platform, Scinode, runs separate agents for research, manufacturing and quality control, choosing production routes and plants and tracking regulatory requirements.1 timeline · 1 person · 1 round · Latest
AlbertSoftware and AI for chemistry and materials R&D at chemical and consumer-goods companies: it captures experiments, structures the data and lets scientists search and reason over it, suggest formulations and design experiments.1 timeline · 3 people · 2 rounds · Latest - Drylab AIAgents that run computational biology workflows end to end (planning the analysis, choosing among more than 800 tools and databases, getting the compute and working through failures), with a separate model of its own checking each analytical decision and flagging method errors; installed on a customer's private infrastructure or workstations.1 timeline · 4 people · 1 round · Latest
- Ineffable IntelligenceA frontier AI lab building a reinforcement-learning 'superlearner' meant to discover knowledge from its own experience, without pre-training on human data.2 timeline · 1 interview · 11 people · 1 round · Latest
- Tianxu Intelligent TechnologyBuilds a weather model for the energy industry (wind speed and solar irradiance rather than general forecasts) with the China Meteorological Administration's public service centre, and is putting the model on satellites of its own to make the observations it trains on.1 timeline · 1 person · 1 round · Latest
Undecimal BioLooks for small-molecule conjugates that carry siRNA drugs into tissues beyond the liver: it screens combinatorial libraries of candidate ligands in cells that report productive delivery, and trains models on the screens. Its first programmes are in kidney and blood-vessel disease.1 timeline · 2 people · 1 round · Latest - Matter MachineA University of Liverpool spin-out pairing AI prediction with experimental and characterisation expertise to find inorganic crystalline materials, aimed at batteries and catalysts.1 timeline · 3 people · 1 round · Latest
BIOTREEA mass-spectrometry service company for metabolomics and proteomics, with its own metabolite standards library. It is moving into single-cell and spatial omics and pitches that data (its MetCell single-cell metabolomics in particular) as training material for virtual-cell models.1 timeline · 2 people · 1 round · Latest - Litai ZhiyaoDesigns peptide drugs with generative models. It runs a platform-plus-pipeline model, taking partner work now and building its own oncology, autoimmune and metabolic programmes; its L2F backers describe the design space as extending to non-natural and mirror-image D-peptides.1 timeline · 4 people · 1 round · Latest
Chai DiscoveryBuilds AI models that predict how molecules bind and design antibodies from scratch, and licenses them to drugmakers as software.5 timeline · 24 people · 4 rounds · Latest - Onyx BioRuns paid, at-home research studies that measure the same people's immune systems over months and years (proteins, single-cell gene expression, metabolites, genetics) beside their clinical histories, processes every sample in its own Berkeley lab, and offers the data to biopharma and AI model developers as a 'virtual immune system'. Its first studies cover autoimmune disease and immune ageing.1 timeline · 2 people · 1 round · Latest
OptoSeekerBuilds lab instruments that move and screen single cells with light (optoelectronic tweezers), for antibody discovery, cell-line development and cell and gene therapy. It pitches the time-stamped single-cell function data they produce as training data for AI biology models.1 timeline · 4 people · 2 rounds · Latest 
Periodic LabsTrains AI models on data from its own autonomous physics and chemistry labs, starting with superconductors and materials for semiconductors.3 timeline · 3 interviews · 10 people · 2 rounds · Latest 
Matchpoint TherapeuticsDesigns covalent small molecules, taken as pills, for autoimmune and inflammatory targets that have been hard to drug. Its ACE platform pairs chemoproteomics with machine learning that guides target choice and compound design.1 timeline · 7 people · 3 rounds · Latest 
RivercellGenerates time-resolved, multimodal single-cell data on how human cells respond to drugs and genetic changes, on its own platform and automated wet lab, to train an 'AI virtual cell' model that predicts those responses.1 timeline · 5 people · 1 round · Latest 
Insilico MedicineGenerative-AI drug discovery (target finding, molecule generation) with its own clinical pipeline led by a fibrosis drug, plus licensing deals with Lilly, Exelixis and others.6 timeline · 1 interview · 7 people · 6 rounds · Latest 
Multiply LabsBuilds robotic clusters that run the manual steps of making cell and gene therapies and other biologics, working with the drugmaker's existing equipment; the drugmakers own and operate the robots.2 timeline · 3 people · 1 round · Latest 
VinciBuilds a physics foundation model with GPU-native solvers that gives engineers solver-accurate thermal, thermo-mechanical and fluid results while a design is still changing, starting with semiconductors.1 timeline · 6 people · 2 rounds · Latest 
WhiteLab GenomicsUses its AI platform, ALFRED, to design the parts of genomic medicines (AAV and non-viral delivery vehicles and synthetic promoters), then tests them in animals, aiming to own and license the validated designs.2 timeline · 5 people · 2 rounds · Latest - Lanyon AIA research lab building an agent that writes scientific solvers and their formal proofs together, so the code provably matches its specification; aimed first at aerospace, propulsion and nuclear simulation.2 timeline · 1 interview · 6 people · 1 round · Latest
Autopoiesis SciencesBuilds AI for scientific reasoning: Aristotle, a research assistant that helps scientists form hypotheses and weigh evidence, and Columbus-1, an autonomous system for long research problems. It says it builds training environments and datasets from expert scientists' judgment.1 timeline · 4 people · 2 rounds · Latest 
ProfluentTrains large protein language models (ProGen3) on its own protein atlas to design gene editors, antibodies and enzymes, and licenses the designs.3 timeline · 8 people · 3 rounds · Latest 
ZenithonWorld models for extreme physics (fusion reactors, rockets and semiconductor manufacturing) meant to replace simulations that take days to months, starting with fusion customers.1 timeline · 1 interview · 2 people · 1 round · Latest 
BiossilUses OpenAI language models to find drug candidates that failed elsewhere and could work in a changed trial design or patient group, then licenses them and runs the new trials.1 timeline · 2 people · 1 round · Latest 
Discovery LoopPublic-benefit company building AI systems that run whole experimental loops at scale, starting with automating machine-learning research itself.2 timeline · 5 people · 2 rounds · Latest 
Prevision MedicineTests a patient's living tumour cells against 100+ drugs and combinations before treatment, with automated microscopy and image analysis, so oncologists can rank drugs by how selectively they kill cancer cells.2 timeline · 4 people · 1 round · Latest - Soge Intelligent ComputingAtomistic AI models with long-range interactions, fast parallel simulation and battery device modelling, sold to battery and chemicals firms; spun out of Shanghai Jiao Tong University.1 timeline · 1 person · 1 round · Latest
Azulene LabsTrains physics-constrained AI chemistry models on quantum-mechanical data, to predict things like drug binding more accurately for pharma and chemicals.1 timeline · 2 people · 1 round · Latest 
LongXun KuangTengSells PWmat, a home-grown first-principles materials simulation package with GPU acceleration and machine-learned force fields, to research groups and battery and chip companies.1 timeline · 2 people · 3 rounds · Latest 
SynMatAIAn agent and four in-house algorithm modules for designing chip-packaging materials, such as optical adhesives, thermal interface materials and photoresists, from lab to production.1 timeline · 2 people · 3 rounds · Latest - Woshi TechnologyA structured chemistry database, AI models for synthesis planning and robotic labs from milligram to kilogram scale, used by chemicals, battery and pharma companies.1 person · 1 round · Latest
Edison ScientificFor-profit spin-out of the FutureHouse nonprofit that sells Kosmos, an AI scientist that reads literature, analyses data and proposes experiments for drug R&D teams.4 timeline · 6 people · 1 round · Latest - Big Picture BioDesigns cancer drug combinations, and their order, with a generative model of how tumours, immune cells and surrounding tissue interact across disease sites and patient groups, then tests the picks in its own wet-lab work; it starts with cancer and immune-mediated diseases.1 timeline · 4 people · 1 round · Latest
- PolyphronGrows human micro-tissues from donor-diverse stem cells in an automated foundry, measures how they respond over time, and trains a tissue model on that data, to test AI-generated drug hypotheses; cardiac tissue first, liver next.2 timeline · 3 people · 1 round · Latest
ChemifyTurns molecule designs into code that robots run (Chemputation) in automated 'Chemifarm' labs, making small molecules for drug and materials partners.3 timeline · 4 people · 4 rounds · Latest - Deep MaterialDesigns metal alloys with its own models and an automated high-throughput lab, then supplies the materials to 3C, aerospace and fusion customers.1 person · 1 round · Latest
Magnitude BiosciencesA Durham University spin-out and contract research lab that tests thousands of compounds on live C. elegans worms, read out by imaging and machine learning.2 timeline · 4 people · 2 rounds · Latest 
DimensoSmart glasses with computer vision and voice that record lab work as it happens, track protocol steps and flag deviations.1 timeline · 4 people · 1 round · Latest - MithrlBuilds a knowledge base from decades of peer-reviewed biomedical studies, keeping what each finding measured and perturbed, and an agentic layer over the frontier models a drug company already uses, so its scientists can judge which hypotheses survive the next experiment; deployed inside the customer's own environment and extended with their data.1 timeline · 3 people · 1 round · Latest
MoleculeMindAI protein structure prediction and design (MoleculeOS, MMFold, MMDesign) sold to drugmakers and industrial-enzyme users, with a few in-house early programs.1 timeline · 15 people · 4 rounds · Latest - RhinovatePairs an AI-scientist agent with robotic labs and pilot scale-up to develop polymers, carbon-fibre composites and battery electrolytes, incubated at Peking University's Shenzhen campus.1 timeline · 3 people · 3 rounds · Latest
Aureka BiotechnologiesDesigns antibodies with its own biological foundation model, AuraIDE (released in open form as OpenDDE), trained and retrained on data from its own single-cell functional screens.3 timeline · 2 people · 1 round · Latest - Superluminal MedicinesDesigns small-molecule drugs for G protein-coupled receptors (GPCRs) with machine learning, structural biology and its own data: cryo-EM, GPCR co-folding models, de novo design and predicted ADME and toxicity. Its lead programme is an MC4R agonist for rare genetic and hypothalamic obesity.1 timeline · 6 people · 2 rounds · Latest
- AIPherTsinghua AIR spin-out building DrugCLIP virtual screening and the PharmAgents decision agent, used to co-develop drug programs with pharma partners; lead ADHD program is pre-PCC.2 timeline · 1 person · 1 round · Latest
InceptiveTrains foundation models on sequence data and its own lab experiments to design RNA and other sequence-based medicines such as mRNA, siRNA and ASOs.2 timeline · 2 people · 2 rounds · Latest 
DataHowHybrid models that mix process knowledge with machine learning to plan and control biopharma manufacturing processes with fewer experiments.2 timeline · 7 people · 2 rounds · Latest 
Tetsuwan ScientificRuns a cloud biology lab where researchers describe an experiment in its ResearchOS software and robots carry it out.1 timeline · 4 people · 1 round · Latest 
TransfyrPuts sensors and multimodal models into labs to record how scientists actually carry out an experiment (their actions, the environment, instrument readings) and turns that record into data for troubleshooting, tech transfer, training and robot instructions. It runs an in-house wet lab that generates its own training data.1 timeline · 6 people · 1 round · Latest - Generalist AIBuilds foundation models for robots, trained on large amounts of real-world physical interaction data, meant to work across robot bodies.1 timeline · 3 people · 2 rounds · Latest
BioptimusBuilds foundation models of biology, starting with pathology images (H-Optimus) and a multimodal model (M-Optimus), and is assembling a large spatial tissue atlas (STELA) to train them.2 timeline · 16 people · 2 rounds · Latest - Unreasonable LabsBuilds a discovery engine that pairs language models with a map of relationships and neurosymbolic mathematical abstractions, to propose and test hypotheses across chemistry, materials science and biology.3 timeline · 1 interview · 4 people · 1 round · Latest
InherentA public benefit corporation building Faraday, a system in which researchers work with self-improving AI agents on open scientific problems.2 timeline · 5 people · 1 round · Latest 
Discovered MaterialsRuns AI agents built on Anthropic models, with physics models it trained, to find new materials for semiconductor chips, starting with thermal materials for data centres and fabs.1 timeline · 6 people · 1 round · Latest 
Orbital IndustriesFormerly Orbital Materials. Uses its atom-simulation model Orb to find materials, then makes and sells the products itself: a PFAS-free data-centre coolant and modular data centres.3 timeline · 5 people · 3 rounds · Latest - Shuye TechMakes smart lab instruments (a voice-controlled electronic pipette, modular lab robots for PCR, ELISA and sequencing prep, and an add-on AI module for older instruments) and pitches them as a networked set of agents that run small, flexible automated labs.1 timeline · 3 people · 1 round · Latest
Relation TherapeuticsFinds drug targets by profiling patients' own tissue with single-cell and genetic data, then machine learning and lab tests; first in osteoporosis.1 timeline · 13 people · 3 rounds · Latest 
AqemiaDesigns small-molecule drug candidates with physics-based calculations feeding a generative model, for its own pipeline and for pharma partners such as Sanofi.1 timeline · 5 people · 3 rounds · Latest 
Latent LabsBuilds generative AI models that design proteins such as antibodies from scratch, and tests the designs in its own wet lab in San Francisco.3 timeline · 9 people · 2 rounds · Latest 
Deep PrincipleGenerative models for chemical reactions and materials (OA-ReactDiff, React-OT), an agent called Mira, and a high-throughput lab, sold to chemicals and consumer-goods companies.3 timeline · 4 people · 7 rounds · Latest 
IktosGenerative design (Makya) and retrosynthesis (Spaya) software joined to robotic synthesis and in-house biology testing, used for partners and its own drug programmes.2 timeline · 6 people · 1 round · Latest 
Alchemab TherapeuticsUses AI over a database of antibodies from people unusually resistant to disease to find protective antibodies and their targets, then develops them as drugs; its first, ATLX-1282 for ALS, is licensed to Eli Lilly.1 timeline · 4 people · 2 rounds · Latest - alqemSearches predicted crystal structures for high-performance materials and makes the shortlist in its own lab, starting with permanent magnets that use no rare earths. Its database work builds on Alexandria, the open materials database its CTO co-built.1 timeline · 4 people · 1 round · Latest
MedraBuilds general-purpose lab robots driven by AI that plan, run and analyse biology experiments on standard instruments, for drug discovery teams.1 timeline · 3 people · 1 round · Latest 
NoetikTrains self-supervised 'virtual cell' models (OCTO) on its own spatial profiling of human tumours to find cancer targets and match drugs to patients.2 timeline · 17 people · 2 rounds · Latest - Jindu ShengkePairs a multi-omics model pretrained on individuals' raw sequencing data (GeneLLM, 1.5B parameters by its own account) with BioFord Agent, a set of agents that design experiments and schedule sequencers, liquid handlers, PCR and mass-spec instruments from different vendors.1 timeline · 1 person · 1 round · Latest
Genesis Molecular AI (formerly Genesis Therapeutics)Builds AI models of how small molecules bind proteins (GEMS platform, Pearl structure model) and uses them for its own and partners' drug programs.2 timeline · 7 people · 2 rounds · Latest - FieldQuantumBuilds a research platform that puts quantum computing, AI models and high-performance computing behind one interface, for calculations in materials, chemistry, biomedicine and manufacturing.1 timeline · 1 person · 1 round · Latest
PercepticSoftware that runs AI agents over a drug company's public, internal and purchased data, with every claim traced to its source, for asset scouting, indication choice and clinical trial design.1 timeline · 5 people · 1 round · Latest - Emmi AISpun out of NXAI's simulation team in Linz to build neural networks that stand in for engineering simulations (fluid flow, heat transfer, particle flows) for aerospace, automotive, energy and semiconductor makers.1 timeline · 7 people · 1 round · Latest
- Recursive SuperintelligenceBuilding AI that improves its own code, harness and training infrastructure through automated experiments, starting with the science of AI and aiming later at fields such as drug discovery, batteries and fusion.1 timeline · 4 people · 1 round · Latest
- Fathom TherapeuticsSimulates protein motion at atomic resolution with quantum chemistry and AI in its Microcosmos engine, and designs small molecules, including degraders, against that dynamic behaviour.1 timeline · 5 people · 1 round · Latest
LabGeniusDesigns multispecific antibodies with EVA, a robotic lab that runs its own experiments and learns from them via machine learning, aiming at solid tumours.2 timeline · 12 people · 4 rounds · Latest 
Axiom BioScreens compounds on live human liver cells at scale and trains AI models on the images to predict drug toxicity before human trials.2 timeline · 10 people · 2 rounds · Latest 
Helixon (Earendil Labs)AI protein modelling plus high-throughput experiments to design antibodies, especially bispecifics, which it licenses to Sanofi and Genentech and develops in-house.5 people · 3 rounds · Latest 
BioMapBuilds xTrimo, large life-science foundation models over protein, DNA and RNA data, and sells models and discovery work to drugmakers such as Sanofi.2 timeline · 3 people · 2 rounds · Latest - AMI LabsA research lab building world models that learn abstract representations of camera and sensor data and predict in that space, for uses it names as industrial process control, robotics, wearables and healthcare.1 timeline · 9 people · 1 round · Latest
CradleSoftware that uses protein language models to suggest better protein variants, which customers test in their own labs; Cradle also runs a wet lab to generate training data.3 timeline · 12 people · 3 rounds · Latest 
StoneWiseAI small-molecule design software and molecule-generation models with its own chemistry lab, sold to drugmakers; its AI-assisted HPK1 inhibitor SWA1211 is cleared for trials.1 timeline · 5 people · 4 rounds · Latest - Ricursive IntelligenceAn AI lab using AI to speed up semiconductor design across the full stack, so that AI models and the chips they run on improve each other.1 timeline · 3 people · 1 round · Latest
Converge BioTrains generative models on DNA, RNA and protein sequences and sells them to drug developers as finished systems for antibody design and optimisation, protein manufacturing yield, and target and biomarker discovery. Customers can fine-tune private copies on their own data.1 timeline · 4 people · 2 rounds · Latest 
OwkinTrains models on multimodal patient data from partner hospitals and licenses an agentic 'AI scientist' (K Pro) plus data to drug companies for target and biomarker work.2 timeline · 16 people · 5 rounds · Latest 
ChemLexRuns an automated chemical synthesis service for drugmakers: AI plans and ranks synthesis routes, and a 24/7 robotic synthesis line makes and records the compounds. It sells this as a CRO service rather than running its own pipeline.1 timeline · 2 people · 1 round · Latest 
EntalpicScreens and generates catalysts and surface materials for industrial chemistry with ML models, atomistic simulation and experimental feedback, lately aimed at atomic layer deposition.2 timeline · 6 people · 1 round · Latest - Accipiter BioDesigns proteins from scratch, with its own computational methods and high-throughput experiments, that combine several mechanisms in one stable molecule, including agonists that activate several pathways on the same cell; its pipeline is in immunology and oncology.1 timeline · 4 people · 1 round · Latest
EvolutionaryScaleBuilt ESM protein language models, including ESM3, which reasons over protein sequence, structure and function to generate new proteins.2 timeline · 5 people · 1 round · Latest - AchiraBuilds molecular simulation models that mix machine learning with physics and quantum chemistry to generate synthetic data for drug design.3 people · 1 round · Latest
Atomic AIPredicts 3D RNA structure with machine learning (ATOM-1) plus in-house wet-lab assays, to design small molecules that hit RNA and RNA-based medicines.1 timeline · 10 people · 2 rounds · Latest 
Emerald Cloud LabRuns remote-controlled robotic labs where scientists write experiments in software and the facility executes and records them.4 people · 1 round · Latest
Institutions and other organizations 64
Universities, national labs and companies the timeline names three times or more.
Flagship Pioneering3 timeline · 7 people · Latest - National University of Singapore4 timeline · Latest
Shanghai Jiao Tong University4 timeline · 1 person · Latest 
SLAC National Accelerator Laboratory3 timeline · 1 person · Latest 
UC Berkeley14 timeline · 10 people · Latest University of Washington11 timeline · 1 interview · 6 people · Latest
- Bristol Myers Squibb3 timeline · Latest
Google Quantum AI6 timeline · 5 people · Latest 
New York University4 timeline · 9 people · Latest 
Imperial College London5 timeline · 6 people · Latest 
MIT14 timeline · 31 people · Latest 
Roche4 timeline · Latest 
University of California San Diego3 timeline · 5 people · Latest 
Ruhr University Bochum3 timeline · 3 people · Latest 
ByteDance5 timeline · 5 people · Latest - King Abdullah University of Science and Technology3 timeline · 4 people · Latest
Peking University10 timeline · 7 people · Latest 
University of Toronto6 timeline · 10 people · Latest 
EPFL4 timeline · 5 people · Latest - Hong Kong University of Science and Technology3 timeline · 1 person · Latest
Moderna3 timeline · 2 people · Latest 
University of Tokyo4 timeline · 2 people · Latest 
Institute of Physics, Chinese Academy of Sciences9 timeline · 13 people · Latest 
NVIDIA12 timeline · 3 people · Latest - Westlake University3 timeline · 2 people · Latest
Accel3 timeline · Latest 
Adaptyv Bio4 timeline · Latest 
Columbia University3 timeline · 4 people · Latest 
Tsinghua University9 timeline · 11 people · Latest 
Eli Lilly6 timeline · Latest 
University of Cambridge8 timeline · 16 people · Latest 
DeepSeek3 timeline · Latest 
Stanford University14 timeline · 31 people · Latest 
Institute of Science Tokyo3 timeline · 4 people · Latest - Macao Polytechnic University3 timeline · Latest
AstraZeneca3 timeline · 2 people · Latest - Mistral AI3 timeline · 2 people · Latest
University of Michigan3 timeline · 2 people · Latest 
Allen Institute for AI3 timeline · 1 person · Latest 
ECMWF4 timeline · Latest 
ETH Zurich5 timeline · 11 people · Latest 
Institute for Protein Design4 timeline · 4 people · Latest Carnegie Mellon University6 timeline · 8 people · Latest
University of Liverpool7 timeline · 6 people · Latest 
GSK4 timeline · 2 people · Latest 
National Hurricane Center3 timeline · Latest 
University College London3 timeline · 1 person · Latest 
Argonne National Laboratory3 timeline · 1 person · Latest 
Cornell University3 timeline · 1 person · Latest 
Johns Hopkins University3 timeline · 2 people · Latest 
National Institutes of Health3 timeline · Latest 
Oak Ridge National Laboratory5 timeline · 1 person · Latest 
US Department of Energy7 timeline · 1 person · Latest 
Meta AI6 timeline · 11 people · Latest 
Mila4 timeline · 2 people · Latest 
Arc Institute4 timeline · 1 person · Latest - University of Oxford3 timeline · 6 people · Latest
- Broad Institute4 timeline · 6 people · Latest
Shanghai AI Laboratory6 timeline · Latest 
Genentech3 timeline · 2 people · Latest 
Novartis3 timeline · Latest 
Technical University of Denmark4 timeline · 3 people · Latest 
EMBL-EBI4 timeline · 1 person · Latest 
PhAI Labs3 timeline · 1 person · Latest