Atlas Bio
AI foundation model for in silico clinical trials to predict drug efficacy and safety.
Website: https://atlas.bio/
Cover Block
Open sources
| Field | Value |
|---|---|
| Name | Atlas Bio |
| Tagline | AI foundation model for in silico clinical trials to predict drug efficacy and safety. |
| Headquarters | Basel, Switzerland [Tech.eu, October 2026] |
| Founded | 2024 [Tech.eu, October 2026] |
| Stage | Seed [Tech.eu, October 2026] |
| Business Model | B2B |
| Industry | Healthtech |
| Technology | AI / Machine Learning |
| Geography | Western Europe |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (2), Simon Romanski and Sebastian Bayer [Tech.eu, October 2026] |
| Funding Label | $10M+ |
| Total Disclosed Funding | ~$12,500,000 [Tech.eu, October 2026] |
Links
Open sources
- Website: https://atlas.bio/
- Careers: https://www.atlasbio.life/careers
- Jobs: https://www.chooseatlas.com/joinus
What an Investor Needs First
PUBLIC Atlas Bio is a Basel-based startup building an AI foundation model for in silico clinical trials, and it is drawing investor attention because it surfaced from stealth in October 2026 with $12.5 million in disclosed funding around a claim that it can predict drug efficacy and safety before Phase II using virtual patients [Tech.eu, October 2026] [Reuters, October 2026] [StartupTicker, October 2026]. Founded in 2024 by Simon Romanski and Sebastian Bayer, the company is positioning itself at the intersection of biotech and machine learning, with a founding story that appears rooted less in academic drug discovery and more in applied AI talent moving into a high-cost, high-failure part of pharma R&D [StartupTicker, October 2026] [Tech Funding News, October 2026].
The product claim is straightforward even if the proof points are still early: Atlas Bio says its model is trained on about two million human RNA readouts and runs roughly 12-week simulations to evaluate patient-level efficacy and safety signals, with the stated aim of reducing animal testing and informing Phase II trial design [The Economic Times, October 2026] [Tech.eu, October 2026] [Tech Funding News, October 2026]. What differentiates the pitch, based on public reporting, is the attempt to move beyond workflow software into decision support for which patients may respond and how trials should be structured, though those claims remain mostly company-described rather than independently validated [Tech Funding News, October 2026] [Tech.eu, October 2026].
The team is one reason the company is getting early coverage. Romanski studied electrical engineering at Karlsruhe Institute of Technology, EPFL, and Stanford, while Bayer previously worked on perception models for autonomous driving and later led Google research on language models used in Search, YouTube, and Google Play; scientific depth appears to be supplemented by Dominik Hartl, whose public biography includes translational medicine and drug development roles across Novartis, Roche, and Galapagos Pharma [Tech Funding News, October 2026] [Granite Bio AG] [LinkedIn, retrieved 2026].
Atlas Bio sells to enterprise pharmaceutical and biotechnology customers under a B2B model, and public sources say it is already working with a top-10 global pharmaceutical company as well as other global pharma firms, although customer names and commercial terms were not disclosed [StartupTicker, October 2026]. Financing is now clearer than commercial traction: the company announced a $10 million seed round led by Creandum, following a previously undisclosed $2.5 million pre-seed led by First Momentum Ventures and LEA Partners, and Creandum partner Katharina Wilhelm joined the board [Tech.eu, October 2026] [StartupTicker, October 2026].
Over the next 12 to 18 months, the key watch items are external validation, repeatability across programs, and evidence that simulated trial outputs translate into measurable development decisions inside pharma workflows. The public case is interesting because Atlas Bio is aiming at a large cost center in drug development with a team that has real machine learning pedigree, but the evidence base still leans heavily on launch-period reporting and company-framed deployment claims rather than disclosed outcome data [Reuters, October 2026] [The Economic Times, October 2026] [Tech.eu, October 2026].
Partially corroborated -- Funding, headquarters, founding date, and board details are corroborated across Tech.eu, Reuters, and StartupTicker, while product performance and customer traction claims rely primarily on launch coverage and company-described reporting.
Taxonomy Snapshot
| Axis | Value |
|---|---|
| Stage | Seed |
| Business Model | B2B |
| Industry / Vertical | Healthtech |
| Technology Type | AI / Machine Learning |
| Geography | Western Europe |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (2) |
| Funding | $10M+ (total disclosed ~$12,500,000) |
Inside the Company
PUBLIC
Atlas Bio is a Basel-based company founded in 2024, according to the company website and company-registration records surfaced via Moneyhouse [Atlas Bio, retrieved 2026] [Moneyhouse, retrieved 2026]. The public positioning is straightforward: the company is building an AI foundation model for in silico clinical trials, aimed at predicting drug efficacy and safety before Phase II [Atlas Bio, retrieved 2026]. The legal officers named in Swiss business records are Simon Romanski as President of the Administrative Board and Sebastian Korbinian Bayer as Vice President of the Administrative Board, each with individual signature authority [Moneyhouse, retrieved 2026].
The chronology is still short, but the early milestones are clear in public materials. The company states it was founded in 2024 [Atlas Bio, retrieved 2026], and by October 2026 it had emerged from stealth with $12.5 million in disclosed financing across a $2.5 million pre-seed and a $10 million seed round, as reflected on Dealroom and on the company site’s public framing around its launch period [Dealroom.co] [Atlas Bio, retrieved 2026]. That sequence matters because Atlas Bio is arriving as a newly capitalized platform company rather than a long-operating services shop, which keeps the story legible but also means the public record remains thin outside formation, positioning, and funding disclosures [Atlas Bio, retrieved 2026] [Dealroom.co].
Partially corroborated -- Confirmed in part by the company website and Moneyhouse, with funding chronology partially corroborated by Dealroom.
Under the Hood
MIXED
The technical claim worth isolating is narrow but consequential: Atlas Bio says it is building an AI foundation model for in silico clinical trials, aimed at predicting drug efficacy and safety before Phase II rather than after animal studies or early human testing have already consumed time and capital [Tech.eu, October 2026]. Public coverage describes the product as running virtual-patient simulations that evaluate patient-level efficacy and safety signals, with the stated goal of reducing animal testing and lowering late-stage trial failure risk [Tech.eu, October 2026] [Reuters, October 2026]. That is an ambitious positioning for a 2024-founded company, and the public record so far supports the existence of the platform and its intended use case more clearly than it supports independent validation of predictive performance [StartupTicker, October 2026].
The reported technical inputs are more specific than the outcome evidence. The Economic Times reported that Atlas Bio's model is trained on about two million human RNA readouts and that simulations run for roughly 12 weeks [The Economic Times, October 2026]. Separate coverage adds that the simulations are meant to identify likely responders, assess multiple diseases simultaneously, and inform Phase II trial design, but those points appear to come from company descriptions relayed in press rather than from a verified public demo or published benchmark study [Tech Funding News, October 2026] [Tech.eu, October 2026].
What can be said with confidence is that Atlas Bio is presenting itself as a model-and-simulation layer for pharma and biotech buyers, not as a broad discovery platform or a services-only consultancy [StartupTicker, October 2026] [Reuters, October 2026]. The public evidence also suggests a translational framing, since the company ties the product to patient stratification, safety readouts, and trial-design decisions rather than to upstream target identification alone [Tech.eu, October 2026] [Tech Funding News, October 2026]. The gap for investors is straightforward: there is still no public benchmark, peer-reviewed validation, or verified demo material in the captured sources that would allow an external reader to test how much of the differentiation rests on proprietary data, model architecture, or workflow integration.
Claim stands unchecked -- This section relies materially on company-described product claims as reported by Tech.eu, Reuters, StartupTicker, Tech Funding News, and The Economic Times, with no independent public benchmark or verified demo in the captured sources.
Market Research
PUBLIC
The market matters now because drug developers are under visible pressure to cut attrition, shorten study design cycles, and produce stronger evidence before expensive mid-stage trials, which is the problem Atlas Bio says it is addressing with virtual-patient simulations before Phase II [Tech.eu, October 2026] [Reuters, October 2026].
The public record here is narrower than investors might like. No named third-party market report in the supplied materials quantifies the TAM, SAM, or SOM for in silico trials, AI-enabled clinical trial design, or pre-Phase II simulation software, so any sizing discussion has to stay anchored to adjacent markets rather than present a false precision. The closest directly relevant public evidence is problem-based rather than revenue-based: Tech.eu frames the backdrop as drug development's "70% failure rate," while Atlas Bio's positioning centers on predicting efficacy and safety before Phase II, where development costs begin to rise sharply and biological uncertainty remains high [Tech.eu, October 2026]. Reuters separately describes the company as running drug trials on virtual patients, which places it in the broader market for software and data infrastructure intended to reduce experimental burden in drug R&D rather than in conventional CRO services alone [Reuters, October 2026].
That framing implies a buyer set concentrated in pharmaceutical and biotechnology R&D organizations, especially translational medicine, early clinical development, and biomarker teams. StartupTicker reports that Atlas Bio is already working with a top-10 global pharmaceutical company and several other global pharma firms, which, if sustained, would suggest demand is coming first from large enterprises that can pay for decision-support tooling tied to trial design and portfolio selection rather than from smaller labs buying point solutions [StartupTicker, October 2026]. The early deployment example, a Basel-based biotech that reportedly removed animal testing from a specific workflow and moved directly to human trials, matters less as proof of scale than as evidence of willingness to test simulation-led workflows in regulated development settings [Tech.eu, October 2026] [StartupTicker, October 2026].
The demand drivers in the cited research are practical. Atlas Bio's model is reported to be trained on about two million human RNA readouts and to run roughly 12-week simulations, with the stated goal of surfacing patient-level efficacy and safety signals before Phase II [The Economic Times, October 2026]. If that capability holds up in practice, the purchasing logic is straightforward: earlier no-go decisions, narrower patient stratification, and potentially fewer wet-lab or animal experiments in selected programs. Tech Funding News adds that the system is meant to identify likely responders, assess multiple diseases simultaneously, and inform Phase II trial design, which points to budget owners who already spend on biomarker discovery, protocol design, and translational analytics rather than on classic EDC or site-management software [Tech Funding News, October 2026].
The adjacent markets are therefore more important than a missing headline TAM number. Atlas Bio appears to sit at the intersection of computational biology, translational medicine software, clinical trial design tools, and parts of the preclinical services stack now handled through animal models and lab experimentation [Tech.eu, October 2026] [The Economic Times, October 2026]. Substitute solutions would include incumbent preclinical testing workflows, in-house biostatistics and biomarker teams, disease-specific simulation tools, and other AI-first platforms trying to model patient response earlier in development [Reuters, October 2026]. That matters because adoption may not require a fully new budget line. In many cases it could come from reallocating spend already earmarked for target validation, preclinical evidence generation, or early clinical planning.
Regulatory and macro forces cut both ways. The favorable side is clear in the coverage: reducing animal testing is a politically and scientifically resonant objective, and software that improves evidence quality before human studies fits broader industry interest in faster, more efficient development programs [Tech.eu, October 2026] [StartupTicker, October 2026]. The limiting factor is equally clear. For a company in this category, commercial expansion depends less on broad AI spending enthusiasm than on whether pharma teams and, eventually, regulators accept simulation outputs as credible enough to shape real trial decisions. None of the supplied sources indicates formal regulatory endorsement, published validation studies, or quantified conversion from pilot work to repeat enterprise adoption, so the market should be read as promising but still early in proof standards [Reuters, October 2026] [The Economic Times, October 2026].
| Cited market signal | Figure | Relevance to demand | Source |
|---|---|---|---|
| Total disclosed Atlas Bio funding | $12.5M | Signals investor willingness to back in silico trial infrastructure | [Tech.eu, October 2026] |
| Pre-seed round | $2.5M | Indicates early institutional support before public launch | [Tech.eu, October 2026] |
| Seed round | $10M | Suggests conviction in enterprise drug-development software angle | [Tech.eu, October 2026] |
| RNA readouts used to train model | Approx. 2,000,000 | Points to data scale as a purchase driver in model credibility | [The Economic Times, October 2026] |
| Simulation duration | Roughly 12 weeks | Frames value proposition around decision speed before Phase II | [The Economic Times, October 2026] |
The table does not size the market directly, but it does show where the early commercial narrative is concentrated: data scale, decision speed, and enough investor appetite to finance a category bet. Until third-party market reports or validated deployment data emerge, the cleaner read is that Atlas Bio is pursuing a budget-rich problem inside pharma R&D rather than a clearly measured standalone software category.
Partially corroborated -- Relies on named public coverage from Tech.eu, Reuters, StartupTicker, Tech Funding News, and The Economic Times, but no third-party market sizing report is present in the supplied sources.
Competition and Substitutes
MIXED Atlas Bio is entering a crowded computational-biology stack, but its public positioning is narrower than many adjacent players: it is selling pre-Phase II, virtual-patient simulation for efficacy and safety prediction rather than a broad AI-for-drug-discovery platform [Tech.eu, October 2026] [Reuters, October 2026].
| Company | Positioning | Stage / Funding | Notable Differentiator | Source |
|---|---|---|---|---|
| Atlas Bio | AI foundation model for in silico clinical trials to predict drug efficacy and safety before Phase II | Founded 2024; total disclosed funding $12.5M | Reported focus on virtual-patient simulations trained on about two million human RNA readouts, with roughly 12-week simulations aimed at patient-level efficacy and safety signals | [Tech.eu, October 2026]; [The Economic Times, October 2026] |
| Noetik | Named competitor in AI-driven biotech | Not established in the provided public source set | Included in the competitive set surfaced in the structured facts | [Structured facts] |
| Unlearn.AI | Named competitor in in silico trial and digital-twin workflows | Not established in the provided public source set | Included in the competitive set surfaced in the structured facts | [Structured facts] |
| Owkin | Named competitor in AI-enabled biopharma infrastructure | Not established in the provided public source set | Included in the competitive set surfaced in the structured facts | [Structured facts] |
The competitive map breaks into three lanes. First are direct in silico trial specialists, where Unlearn.AI Third are the incumbent substitutes, conventional animal models, CRO-led trial design, and internal biostatistics teams, which remain the default buying behavior Atlas Bio is trying to displace rather than software peers alone [Tech.eu, October 2026] [StartupTicker, October 2026].
Atlas Bio's most defensible edge today appears to be the combination of technical pedigree and a specific dataset claim, though both still need harder external validation. Public coverage says the model is trained on approximately two million human RNA readouts and that the founders come from autonomous-driving and large-scale model backgrounds, while scientific leadership adds drug-development experience from large biopharma settings [The Economic Times, October 2026] [Tech Funding News, October 2026] [Granite Bio AG]. That mix matters because enterprise pharma buyers usually want both model-building credibility and translational-science fluency. The durability of that edge is less certain. Talent can move, model techniques diffuse, and a dataset claim without public evidence on exclusivity or prospective clinical validation is perishable if larger rivals or incumbents can assemble comparable multimodal training corpora [The Economic Times, October 2026].
The company is most exposed where competitors can offer either deeper clinical validation or broader workflow ownership. Unlearn.AI, Atlas Bio also does not appear, from public reporting here, to own a proprietary commercial channel such as a major pharma distribution partnership, a CRO channel, or a regulator-linked program, which means go-to-market may depend on direct enterprise selling and early reference accounts that remain mostly unnamed in public [StartupTicker, October 2026] [Tech.eu, October 2026].
The most plausible 18-month scenario is a sorting of the field by evidence quality rather than by model claims alone. Atlas Bio is the likely winner if it can turn its early reported pharma work into named validation studies, repeatable proof that simulated outputs improve Phase II design, and at least a small set of referenceable enterprise deployments [StartupTicker, October 2026] [Tech.eu, October 2026]. In that case, its narrow focus could help it become a specialist layer inside existing pharma workflows. Noetik and Unlearn.AI
Reasoned from indirect evidence -- Core Atlas Bio facts are corroborated by Tech.eu, Reuters, StartupTicker, The Economic Times, and Tech Funding News, but competitor-specific comparisons rely partly on structured-fact naming without equivalent source depth for each peer.
Opportunity
PUBLIC
If Atlas Bio executes, the prize is not a niche software outcome but a position inside the core decision stack of drug development, where one model influences whether a program advances, which patients it targets, and how sponsors design Phase II studies [Tech.eu, October 2026] [StartupTicker, October 2026].
The headline opportunity is straightforward: become a decision-support platform that pharma teams use before committing capital to mid-stage clinical development. That is a meaningful wedge because Atlas Bio is not presenting itself as general-purpose research software, but as an in silico trial system aimed at predicting efficacy and safety before Phase II, a point in the pipeline where mistakes become expensive and hard to reverse [Tech.eu, October 2026] [Reuters, October 2026]. The public evidence remains early and partly company-described, but it is enough to make the upside legible. The company has disclosed $12.5 million in financing, including a $10 million seed led by Creandum, and reported work with a top-10 global pharmaceutical company plus other pharma firms, which suggests the product is already being evaluated in the buying center that matters most [Tech.eu, October 2026] [StartupTicker, October 2026].
A second reason the opportunity is reachable rather than purely conceptual is team-market fit. Atlas Bio was founded by Simon Romanski and Sebastian Bayer, whose backgrounds sit in machine learning and autonomous systems, while scientific strategy leader Dominik Hartl is described in public sources as having more than 10 years of drug-development experience and prior roles at Novartis, Roche, and Galapagos Pharma [Tech Funding News, October 2026] [Granite Bio AG] [StartupTicker, October 2026]. That combination does not prove commercial success, but it does fit the problem the company says it is solving: a model-heavy product that still has to map onto real clinical development workflows [Tech.eu, October 2026].
| Scenario | What happens | Catalyst | Why it's plausible |
|---|---|---|---|
| Pre-Phase II standard layer | Atlas Bio becomes a recurring software and services platform used by large pharma teams to screen programs before Phase II and shape trial design. | A named top-tier pharma deployment converts into broader internal adoption across multiple indications [StartupTicker, October 2026]. | The company already reports work with a top-10 global pharmaceutical company, and its product is positioned specifically around efficacy, safety, and Phase II design rather than broad discovery tooling [StartupTicker, October 2026] [Tech.eu, October 2026]. |
| Animal-testing displacement wedge | Atlas Bio wins adoption first where sponsors want to reduce preclinical animal work and move faster into human studies. | Additional case studies showing simulation output can replace parts of existing animal-testing workflows [Tech.eu, October 2026]. | Public coverage cites one Basel-based biotech that reportedly eliminated animal testing in a specific workflow and proceeded directly to human trials using Atlas Bio's simulation output, which at minimum indicates a concrete use case around workflow substitution [Tech.eu, October 2026] [StartupTicker, October 2026]. |
| Multidisease simulation platform | Atlas Bio expands from single-study support into a broader platform used across therapeutic areas, increasing value per customer over time. | Validation that the same foundation model can support multiple disease programs and patient stratification tasks [Tech Funding News, October 2026]. | The company says its simulations can identify likely responders, assess multiple diseases simultaneously, and inform Phase II trial design, implying a platform model rather than a one-off point solution [Tech Funding News, October 2026] [The Economic Times, October 2026]. |
What compounding looks like here is a data-and-workflow flywheel. Atlas Bio says its foundation model is trained on about two million human RNA readouts, and the product is meant to generate patient-level efficacy and safety signals over roughly 12-week simulations [The Economic Times, October 2026] [Tech.eu, October 2026]. If customers use the system repeatedly across programs, Atlas Bio could gain three advantages at once: more validated use cases by indication, deeper embedding in clinical planning workflows, and higher switching costs as outputs begin to influence protocol design and go or no-go decisions. The public record does not yet show that this flywheel is fully active, but reported work with a top-10 pharma company and several other global pharma firms is the clearest sign that early distribution into enterprise biopharma may already be starting [StartupTicker, October 2026].
The size of the win is best framed through strategic value, not current fundamentals. There is no confirmed market-sizing dataset in the public materials, and no disclosed revenue base to anchor a valuation model. Still, if Atlas Bio were to become a trusted pre-Phase II decision layer for large drug developers, the outcome could resemble a high-value clinical software and AI platform rather than a narrowly scoped biotech tool. With $12.5 million in disclosed funding and board support from Creandum through partner Katharina Wilhelm, the company has at least enough backing to pursue that platform path in earnest [Tech.eu, October 2026]. On that scenario, Atlas Bio could plausibly support a venture-scale outcome in the hundreds of millions of dollars in enterprise value (scenario, not a forecast), especially if it converts early pharma engagements into repeatable deployment before larger incumbents define the category.
Partially corroborated -- This section relies on multiple named public sources for funding, founders, and reported customer activity, but several material product and traction claims remain company-described or reported by a single outlet, including the top-10 pharma engagement and animal-testing replacement example.
Sources
Open sources
[Tech.eu, October 2026] Swiss startup Atlas Bio secures $12.5M to tackle drug development’s 70% failure rate and move beyond animal testing | https://tech.eu/2026/10/27/swiss-startup-atlas-bio-secures-125m-to-tackle-drug-developments-70-failure-rate-and-move-beyond-animal-testing//
[Reuters, October 2026] Swiss AI drug trial start-up receives funding from Spotify-backer Creandum | https://whtc.com/2026/10/27/swiss-ai-drug-trial-start-up-receives-funding-from-spotify-backer-creandum/
[StartupTicker, October 2026] Atlas Bio steps out of stealth with $12.5m funding | https://www.startupticker.ch/en/news/atlas-bio-steps-out-of-stealth-with-12-5m-funding
[Tech Funding News, October 2026] Ex-Google, self-driving engineers’ Atlas Bio raises $12.5M run drug trials as AI simulations | https://techfundingnews.com/atlas-bio-12-5m-ai-simulate-phase-ii-drug-trials/
[The Economic Times, October 2026] Swiss AI drug trial startup receives funding from Spotify-backer Creandum | https://economictimes.com/tech/artificial-intelligence/swiss-ai-drug-trial-startup-receives-funding-from-spotify-backer-creandum/articleshow/135238709.cms
[Atlas Bio, retrieved 2026] Atlas Bio | https://atlas.bio/
[Dealroom.co] Atlas Bio company information, funding & investors | https://app.dealroom.co/companies/atlas_bio_1
[Granite Bio AG] Dominik Hartl, MD | https://www.granite.bio/team/dominik-hartl
[LinkedIn, retrieved 2026] Dominik Hartl, MD | LinkedIn | https://www.linkedin.com/in/dominik-hartl-md-5a2b3bb4/
Articles about Atlas Bio
- Atlas Bio Replaces Animal Testing With 12-Week AI Simulations — The Swiss startup has convinced Creandum to lead a $10 million seed round for its AI model, trained on two million RNA readouts.