What Edison Scientific does

Edison Scientific, Inc. is a Delaware-incorporated, privately held AI company headquartered in San Francisco that builds Kosmos, an "AI Scientist" platform designed to execute end-to-end biopharma R&D workflows — from target identification through clinical development and regulatory filing — autonomously. Kosmos is built on structured scientific "world models" that maintain coherence across tens of millions of tokens and is composed of specialized sub-agents: Edison Literature (a PaperQA3-based deep research agent reading figures, tables, and formulas across 150M+ papers, patents, and clinical trials), Edison Analysis (a Jupyter-native scientific coding agent executing Python/R/Bash), Edison Molecules (a chemistry agent with 30+ tools for retrosynthesis, ADMET, and toxicity), and Edison Precedent. Training and fine-tuning use the open-source Aviary framework integrated with NVIDIA NeMo Gym and NeMo RL, and the platform is model-agnostic. The company is the 2025 commercial spinout of the FutureHouse non-profit, co-founded by CEO Sam Rodriques and CTO Andrew White, who originated ChemCrow, PaperQA, Aviary, and ether0.

Edison monetizes through a mix of usage-based pay-as-you-go credits (Kosmos launched at $200 per run, 200 credits at $1/credit), subscription plans with monthly credit allocations plus non-expiring add-on credits, a freemium academic tier via the Edison Playground (10 free credits/month plus a one-time 2,000-credit block for existing academics), and quote-based Master Services Agreements for enterprise biopharma deployments. Enterprise contracts bundle SOC 2 Type II-compliant on-prem deployment, forward-deployed scientists, and custom world models fine-tuned via RL and supervised fine-tuning on partner data; the flagship example is the strategic collaboration with Incyte Corporation covering target discovery, target validation, and translational biology. Distribution is hybrid: enterprise field sales to biopharma R&D teams, a self-serve platform at platform.edisonscientific.com, a public Python API/SDK, and embedded workflow integrations (Slack, Teams, email, ELNs).

The company closed a $70 million seed round in December 2025 co-led by Spark Capital and Triatomic Capital (with a related SEC Form D filing for ~$68.89M), holds SOC 2 Type II certification, and operates with a small founding leadership team of ~8-9 named executives and members of technical staff. Edison also authors and maintains the scientific AI community's de facto evaluation benchmarks (LABBench2, BixBench, BixBench-Hypothesis, ether0), publishes open-source frameworks (PaperQA, Aviary, hypotest), and has been featured at NVIDIA GTC 2026, in CB Insights' Early-Stage AI Trends Report, and in tier-one press including Bloomberg and Forbes.

Name

Edison Scientific

Legal name

Edison Scientific, Inc.

Website

https://edisonscientific.com

Company type

Private

Founded year

2025

Operating status

Operating

Headcount range

51–100 employees

Short description

Edison Scientific is an AI company building Kosmos, an autonomous AI Scientist platform that executes end-to-end biopharma R&D workflows from target discovery through clinical development, serving biopharma enterprise customers, academic researchers, and institutions via subscriptions, usage-based credits, and enterprise MSAs.

Ownership category

akta.pro rank

Product category

AI Drug Discovery Platform

NAICS

Scientific Research and Development Services (5417), Research and Development in Biotechnology (except Nanobiotechnology) (541714)

SIC

Services-Prepackaged Software (7372)

akta.pro primary industry

AI/ML Platforms for Drug Discovery & Experiment Optimization (HLAGAJAO)

akta.pro secondary industry

Knowledge Management & Scientific Collaboration / Literature Intelligence (HLAGAJAN)

Keywords

• AI drug discovery
• Autonomous scientific research
• Biopharma R&D platform
• AI scientist agents
• Computational biology tools

Headquarters

HQ city

San Francisco

HQ country

United States

HQ region

North America

Offices2 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Personnel, Technology or R&D, Operations, Marketing or Sales, Infrastructure

Revenue model

1. Subscription plans (Monthly Plan Credits): Recurring subscription Plans with a defined Initial Term and automatic Renewal Terms. Plans come with a base level of Monthly Plan Credits that reset at the start of every Renewal Term; unused credits expire. Plans are specified on the public pricing page.
2. Usage-based pay-as-you-go credits (per-run): Kosmos launched at $200 per run (200 credits at $1/credit), with Founding Subscriptions locking in the $1/credit price. Add-On Credits can be purchased separately, do not reset or expire, and are used after Monthly Plan Credits are exhausted.
3. Enterprise Master Services Agreements (MSA): Enterprise contracts governed by MSAs and Data Processing Agreements (DPAs) for biopharma R&D deployments; example includes the Incyte strategic collaboration embedding Kosmos into discovery and development. Forward-deployed scientists and custom world models are part of the enterprise offering.
4. Freemium / academic tier: Edison Playground provides Literature, Analysis, and Precedent agents with 10 free credits per month; existing academic subscribers receive a one-time 2,000-credit block grant (in lieu of the prior 650/month), supporting a long-term freemium funnel for the academic community.

Pricing tiers

Go-to-market motion4 records

Distribution channels5 records

Marketing channels7 records

Core offering

Edison Scientific offers an AI Scientist platform called Kosmos that autonomously executes end-to-end biopharma R&D workflows from target identification through clinical development and filing, by orchestrating specialized agents for literature research (Edison Literature on PaperQA3), scientific data analysis (Edison Analysis in Jupyter notebooks), and chemistry reasoning (Edison Molecules with 30+ tools). The platform is delivered through enterprise Master Services Agreements with on-prem SOC 2 Type II deployment and forward-deployed scientists, alongside a self-serve web platform with monthly subscription credits, pay-as-you-go usage ($200/run, $1/credit), a Python API, and a freemium Edison Playground tier for academic researchers.

Product overview

Edison Scientific offers a platform-plus-modules AI architecture for scientific R&D, built around Kosmos — its flagship AI Scientist for biopharma R&D that orchestrates end-to-end workflows from hypothesis generation to clinical development. The platform exposes three specialized agents that can be used individually or composed by Kosmos: Edison Literature (a deep research agent built on PaperQA3 that searches over 150M+ papers, patents, and clinical trials), Edison Analysis (a Jupyter-native data analysis agent that executes Python/R/Bash and integrates with the top 20 biomedical databases via its search_and_retrieve tool), and Edison Molecules (a chemistry agent with 30+ specialized tools covering retrosynthesis, ADMET, and toxicity). The commercial platform is complemented by the Edison Playground, a free-tier research environment that hosts Literature, Analysis, and Precedent agents. Underpinning the platform are Edison's open-source research artifacts — the PaperQA3 framework, the Aviary RL training framework, the ether0 chemical reasoning model, and the LABBench2, BixBench, and BixBench-Hypothesis benchmarks.

Differentiator

Problem solved

Functional benefit

Brands

• Kosmos: Edison's flagship AI Scientist platform for R&D teams; performs autonomous scientific investigations across the full biopharma R&D lifecycle, from hypothesis generation to clinical development.
• Edison Analysis
• Edison Literature
• Edison Playground
• Edison Labs

Products and services

• Kosmos AI Scientist Edison's flagship AI Scientist agent that autonomously executes end-to-end biopharma R&D workflows from hypothesis to registration for biopharma R&D teams. Reads literature, analyzes data, generates hypotheses, steers investigations mid-run, branches hypotheses, and integrates with Slack, Teams, and email for synchronous collaboration.
• New Kosmos (Biopharma R&D Edition) Rebuilt Kosmos platform covering the full biopharma R&D lifecycle from target identification through clinical development and filing, for biopharma R&D teams and forward-deployed enterprise deployments.
• Edison Literature Deep research and information retrieval agent that searches and synthesizes over 175M+ scientific articles, patents, and clinical trials with multimodal figure and table understanding. Powers the literature search component of Kosmos and is available individually via the platform and API.
• Edison Analysis Science-native data analysis agent that iteratively updates Jupyter notebooks with Python, R, and Bash in a Docker environment; integrates search_and_retrieve specialist connectors to ~20 priority scientific data sources including Zenodo, PDB, ClinicalTrials.gov, GEO, CCLE, and GDSC. Reaches 46% on BixBench and 70% on Biomni Eval1.
• Edison Molecules Chemistry-specialized agent that unifies synthesis planning, molecular property prediction, safety assessment, and chemical data retrieval via 30+ specialized chemistry tools including retrosynthesis planners, ADMET predictors, toxicity models, and interfaces to ChEMBL, PubChem, and Chem-space.
• Edison Playground Public free-tier access to Edison's core research agents—Literature, Analysis, and Precedent—with 10 free credits per month for all users. Provides an entry point for academics and individual researchers to use Edison's agents.
• Enterprise Kosmos Deployment (MSA) Enterprise deployment of Kosmos under a Master Services Agreement and Data Processing Agreement, including forward-deployed scientists, custom world models built on partner data, on-prem SOC 2 Type II deployment, full governance and auditability, and trainable architecture supporting RL and supervised fine-tuning on partner data.
• Edison Client API & SDK Python Edison Client SDK and API providing programmatic access to Edison's agents (literature-20260216, literature-20260216-high, Edison Analysis, Edison Molecules, and Kosmos) for enterprise and academic integration, documented in the Edison Cookbook.

Quantifiable outcome

• Kosmos run = ~6 months of human scientist work (beta user estimate, 6.14-month average across 7 scientists for 20-step runs)
• +8 more outcomes

Named customers4 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration14 records

AI capability13 records

Feature10 records

Partnerships

Nine partnerships are on record, tiered industry ecosystem / co-adoption by incyte, flagship enterprise partner, core infrastructure partner, research collaboration, foundational origin / continuing collaboration, operational vendor, board / governance connection and academic research collaboration.

• Genesis Molecular AIindustry ecosystem / co-adoption by incyteOthers · 24 May 2026Incyte expanded its AI drug discovery capabilities through parallel partnerships with Genesis Molecular AI and Edison Scientific, integrating both platforms into its R&D workflows. Edison is co-listed with Genesis as a partner of Incyte (not a direct Edison partner).
• Incyte Corporationflagship enterprise partnerStrategic or Co-development Partner · 19 May 2026Strategic collaboration to deploy the Kosmos AI platform across Incyte's drug discovery and development lifecycle, initially focused on target discovery, target validation, and translational biology. Both teams will measure impact on decision quality and pipeline productivity as the deployment expands across Incyte's broader R&D organization. Patrick Mayes, EVP & CSO of Incyte, provided a quote.
• NVIDIAcore infrastructure partnerTechnology or Integration · 16 March 2026Deep technology partnership across NVIDIA NeMo (Gym, RL), Nemotron 3, and Nemotron Parse. Edison trains Kosmos agents using Aviary + NeMo Gym/RL with GRPO and FP8 RL on Nemotron models; Nemotron Parse powers Edison Literature's figure/table extraction. NVIDIA published a case study and tech blog on the integration. BixBench-Hypothesis and BBH-Train were co-developed with NVIDIA.
• UC Berkeleyresearch collaborationStrategic or Co-development Partner · 17 February 2026Co-developed LABBench2 (1,900-task benchmark for AI in scientific work) alongside FutureHouse and the Broad Institute. LABBench2 is published on Hugging Face (futurehouse/labbench2) with the harness on GitHub.
• Broad Instituteresearch collaborationStrategic or Co-development Partner · 17 February 2026Co-developed LABBench2 with Edison Scientific, UC Berkeley, and FutureHouse. The benchmark covers 1,892 tasks across 11 categories for evaluating AI systems in scientific work.
• FutureHousefoundational origin / continuing collaborationStrategic or Co-development Partner · 5 November 2025Edison Scientific is the commercial spinout of the FutureHouse non-profit. Founders Sam Rodriques and Andrew White co-run both organizations with conflict-of-interest controls. FutureHouse continues to focus on basic research and educational access, while Edison scales the AI Scientist commercially. Edison inherited the Robin, PaperQA, ChemCrow, ether0, and Aviary projects from FutureHouse.
• Stripe, Inc.operational vendorOthersProcesses all Edison subscription and credit payments through a PCI-DSS-compliant flow. Edison Scientific does not store full payment card information; Stripe shares limited transactional data (payment status, billing country) with Edison.
• Tony Kulesa / Pillar VCboard / governance connectionStrategic or Co-development PartnerTony Kulesa, Partner at Pillar VC and a founding board member of FutureHouse, is listed as Executive Director on Edison's team page, maintaining the link between Edison and Pillar VC's biotech/deep tech portfolio (Latent Labs, Nabla Bio, Portal Biotech, Amber Bio, Rowan Science).
• University of Rochesteracademic research collaborationStrategic or Co-development PartnerTonio Buonassisi, Professor of Mechanical Engineering at MIT (MIT logo shown on Edison's homepage as a trusted partner), provided a public testimonial about Kosmos executing ~292 independent analysis trajectories in 24 hours. The article also references University of Rochester-led LLM-driven materials discovery work as part of the broader research ecosystem Edison operates in.

Scale indicators11 records

Recent moves7 records

Expansion highlights6 records

Direct peers

• Absci: Generative AI drug creation company combining AI denovo design with a proprietary wet-lab platform to deliver novel antibody therapeutics. Comparable to Edison in merging AI-driven design with end-to-end biopharma execution.
• Owkin: AI-biotech applying ML to patient-derived data for drug discovery, translational research, and clinical development. Comparable to Edison as an AI platform targeting the biopharma R&D-to-clinical lifecycle with research-grade infrastructure.
• Exscientia: AI-driven precision medicine company using AI to design and develop novel therapeutics, with end-to-end discovery-to-clinical capabilities. Comparable to Edison's Kosmos as a full-lifecycle AI drug discovery platform.
• Insitro: ML-driven drug discovery company combining generative biology, data generation, and ML models to accelerate therapeutic discovery. Compares to Edison in its ambition to bring AI into the biopharma R&D loop end-to-end.
• Recursion Pharmaceuticals: AI-native biopharma combining high-throughput wet-lab experimentation with ML models to discover and develop new therapeutics. Directly comparable to Edison as a vertically integrated AI-for-drug-discovery platform targeting the full R&D lifecycle.
• Genesis Molecular AI: Generative and predictive AI platform for small-molecule drug discovery; co-listed as an Incyte partner alongside Edison. Compares to Edison's Molecules agent and Kosmos's chemistry reasoning capabilities.

Broad incumbents

• Isomorphic Labs: Alphabet-backed commercial drug discovery spinout from DeepMind building an AI-first drug design engine. Comparable to Edison as a well-capitalized, full-lifecycle AI drug discovery effort, though with broader financial and compute resources.

Others

• Benchling: Cloud platform for scientific R&D data management (ELNs, sample tracking, molecular biology). Adjacent rather than directly competitive: Benchling manages the laboratory data that Edison's Kosmos agents analyze and integrate via its search_and_retrieve tool.

Emerging players

• Latent Labs: AI-first drug design company focused on programmable biology, backed by Pillar VC (the same VC whose partner sits on Edison's board). Comparable as a frontier AI-for-biology emerging player targeting biopharma R&D acceleration.
• Chai Discovery: AI platform focused on molecular structure prediction and protein design for drug discovery; featured alongside Edison on NVIDIA's Bio International Convention agentic-AI panel. Closely comparable in scope and emerging-player profile.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks6 records

Key highlights7 records

Customer concentration

• GitHubgithub.com/EdisonScientific/hypotest

Compliance1 record

Revenue estimate

Valuation estimate

Number of profiles

Profiles9 records

Funding overview

Funding rounds2 records

Investors7 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Edison Scientific