Science AI apps

AI R&D and innovation intelligence over patents and research.

An enterprise R&D and innovation-intelligence platform that aggregates patents, scientific literature, market data, and grants into a searchable AI workspace. Built for research and IP teams doing technology scouting, competitive analysis, freedom-to-operate assessments, and landscape reports.

NVIDIA's development platform for AI drug discovery.

BioNeMo is NVIDIA's development platform for AI-driven biology and drug discovery. It bundles open biology foundation models, libraries, datasets, and GPU-optimized training recipes, plus NVIDIA NIM microservices that package models for protein structure prediction, molecular design, and virtual screening. The BioNeMo Framework is open source; production models deploy as managed microservices.

Open agentic research assistant for scientists, from Ai2.

Asta is an open ecosystem of AI agents for science from the Allen Institute for AI (Ai2). The Asta assistant helps researchers find relevant papers, generate cited literature summaries, and (in beta) run basic data analyses, with planned hypothesis generation and experiment replication. It ships alongside AstaBench, a benchmark suite for evaluating scientific agents.

Agentic AI for life-sciences R&D, grounded in a biomedical knowledge graph.

Causaly is an agentic AI platform for life-sciences research and development that distills tens of millions of scientific publications and datasets into a high-precision biomedical knowledge graph, then lets domain-specific AI agents plan, search, reason, and synthesize evidence. Researchers use it to map biomarkers, molecular targets, and disease mechanisms and to move from hypothesis to evidence-backed decision in hours rather than weeks. It is sold to pharmaceutical and life-sciences enterprises.

A biodiversity-scale protein database for training biological AI models.

A techbio company building BaseData, a proprietary database of protein and genome sequences collected from biodiversity worldwide under formal benefit-sharing agreements. The dataset is designed to train generative AI foundation models for protein and enzyme design, powering drug discovery, industrial biology and planetary-health applications that public databases can't reach.

AI platform that designs new materials on demand, like a search engine for molecules.

CuspAI is a materials-discovery company building an AI platform that takes a target set of properties and generates viable, synthesizable material candidates, with carbon capture and other sustainability applications as early focuses. It combines generative AI foundation models, deep learning, and molecular simulation into one loop to shortcut conventional trial-and-error discovery. It is engaged as an enterprise platform and research partner rather than a self-serve product.

An agentic platform for enterprise scientific R&D on Azure.

An enterprise platform that lets research teams deploy specialized AI agents over a graph-based knowledge engine to run the full R&D loop — exploring literature, generating hypotheses, optimizing experiments, and validating results. It spans chemistry, materials science, biology, pharmaceuticals, and semiconductors, and runs on Azure HPC and AI Foundry infrastructure with enterprise governance. Researchers can integrate their own models, tools, and datasets.

AI platform for preclinical drug-discovery research.

BenchSci builds AI software for preclinical pharmaceutical R&D, helping scientists generate hypotheses, identify targets, and design experiments faster. Its platform reads and structures public scientific literature plus a customer's proprietary data into a Biological Evidence Knowledge Graph that scientists query in natural language. Its newer EMET agentic research environment chains domain-specific scientific tools, alongside the established ASCEND disease-biology platform used across major pharma.

AI-driven drug discovery on the Recursion OS platform.

Recursion is a clinical-stage, technology-enabled drug discovery company built around Recursion OS, which pairs one of biopharma's largest proprietary datasets with automated wet labs and machine learning for target and molecule discovery. Its LOWE interface lets scientists query the platform in natural language to find targets, generate compounds, and schedule experiments. It advances an internal pipeline and partners with pharma rather than selling self-serve software.

Generative AI drug discovery — the end-to-end Pharma.ai suite.

Insilico Medicine is a clinical-stage drug discovery company whose Pharma.ai suite applies generative AI across the pipeline: PandaOmics for target identification, Chemistry42 for de novo molecule design, and inClinico for clinical-trial outcome prediction. The cloud software is licensed to pharma, and the company says it serves 13 of the top 20 global drugmakers. It also runs its own internal pipeline of AI-discovered drug candidates.

Autonomous AI research lab whose agents run experiments end-to-end.

Autoscience is an applied research lab building a virtual laboratory of autonomous AI scientists that generate hypotheses, design and run machine-learning experiments, and write up the results. Its agent Carl authored a full-length paper that passed double-blind peer review with only minor human edits. The San Mateo company raised a $14M seed round in March 2026 led by General Catalyst.

AI co-scientist for designing and running experiments.

Potato is an AI platform for life-science researchers that pairs literature search and synthesis with an AI co-scientist, 'Tater,' to plan and run experiments end to end. It can generate hypotheses and protocols, translate them into robotic scripts, and analyze results, with human review at each step. It's used in academic and biotech labs and offers free and paid plans.

Browser-based generative AI for push-button protein design.

Latent Labs builds frontier generative AI models for molecular biology, delivered through a no-code web platform. Its Latent-X model designs de novo protein binders — mini-binders and macrocycles — directly in the browser, letting scientists upload a target and generate candidates without specialized infrastructure. The platform is in early access.

AI scientists paired with autonomous labs to discover new materials.

Periodic Labs is building AI scientists and the autonomous laboratories for them to operate, automating hypothesis generation, experiments, and analysis across the physical sciences. Its initial targets include discovering higher-temperature superconductors and new materials for chipmakers. It is a research-stage company deploying with industry partners rather than a self-serve product.

AI agents that automate scientific literature research and discovery.

FutureHouse is a nonprofit AI-for-science lab building AI agents that automate steps of the research process. Its free FutureHouse Platform exposes specialized agents — including Crow, Falcon, Owl, and Phoenix — for literature search, synthesis, and chemistry, plus Kosmos, an 'AI Scientist' that runs autonomous multi-step analyses. The agents are available through a web app and an open API.

Autonomous AI labs pursuing scientific superintelligence.

Lila Sciences is building a 'scientific superintelligence' platform that pairs a proprietary AI model with autonomous AI Science Factory lab instruments to generate hypotheses, design and run experiments, and learn in real time. It offers Catalyst, which augments existing R&D, and Creation, for new discoveries, across materials, therapeutics, chemicals, and energy. It is a platform and research company engaged through partnership rather than self-serve.

Frontier protein language models (the ESM family) for biology.

EvolutionaryScale builds frontier protein language models — the ESM family, including ESM3, which reasons jointly over sequence, structure, and function. ESM3-open ships open weights and code under a non-commercial license, while larger models are served via its Forge API and AWS/NVIDIA partnerships. It is both an open-model provider and an API platform for protein design and understanding.

Generative protein language models that design novel proteins.

Profluent uses generative AI — its ProGen-lineage protein language models — to author novel proteins from scratch or from natural scaffolds, for medicine and agriculture. It open-released and published OpenCRISPR-1 in Nature, and is expanding into antibodies and enzymes. It operates as an AI protein-design company working through partnerships with open-sequence releases, not a self-serve consumer platform.

AI protein-engineering platform for faster lab design rounds.

Cradle is an AI-powered protein-engineering SaaS platform that guides scientists through design rounds — candidate generation, multi-property optimization, and results analysis — using ML models trained per-customer on proprietary wet-lab data. It runs its own validation wet lab in Amsterdam and serves biopharma and industrial-biotech R&D teams including Novo Nordisk and Johnson & Johnson. It is sold via subscription with no royalties, and customers keep full IP.

Foundation models for molecular structure and antibody design.

Chai Discovery builds foundation models for drug discovery — Chai-1 for biomolecular structure prediction across proteins, ligands, DNA/RNA, and glycans, and Chai-2 for de novo antibody design. Chai-1 weights and inference code are released for non-commercial use, and a free web server runs the model in-browser. It is both a model provider and a usable platform.

AI-first drug design built on AlphaFold for pharma R&D.

Isomorphic Labs is an AI-first drug discovery company spun out of Google DeepMind, building generative and predictive ML models for molecular design and drug-performance prediction. Its Drug Design Engine builds on AlphaFold to advance internal and partnered pipelines with Eli Lilly, Novartis, and J&J. It is a drug-discovery company that works through pharma R&D partnerships rather than a self-serve product.

AI deep-research agent that finds every relevant paper on a scientific question.

A research-discovery platform that takes a natural-language question and autonomously searches, reads, and evaluates hundreds of academic papers — following citation trails and explaining each result's relevance, mimicking how an expert runs a literature review. Built by MIT quantum-physics PhDs and backed by Y Combinator; used by researchers at MIT, Harvard, and Caltech.