Mantis Biotech
Builds digital twins of humans using synthetic datasets for medicine, sports, and defense.
Website: https://mantisbiotech.com
Cover Block
PUBLIC
| Field | Value |
|---|---|
| Name | Mantis Biotech |
| Tagline | Builds digital twins of humans using synthetic datasets for medicine, sports, and defense |
| Headquarters | San Francisco, California |
| Founded | 2025 |
| Stage | Seed |
| Business Model | B2B |
| Industry | Healthtech |
| Technology Type | Biotech / Life Sciences (synthetic data, AI) |
| Geography | North America |
| Growth Profile | Venture Scale |
| Founding Team | Solo Founder |
| Funding Label | Seed |
| Total Disclosed | ~$7.4M [The AI Insider, April 2026] |
Links
PUBLIC
- Website: https://mantisbiotech.com/
- LinkedIn: https://www.linkedin.com/company/mantis-biotech
- Y Combinator profile: https://www.ycombinator.com/companies/mantis
- Crunchbase: https://www.crunchbase.com/organization/mantis-biotech
Executive Summary
PUBLIC
Mantis Biotech is a San Francisco company building synthetic datasets that power digital twins of the human body, with target use cases spanning clinical trial design, sports performance, and defense [TechCrunch, March 2026]. The company emerged from Y Combinator and disclosed approximately $7.4M in seed funding in early 2026, with participation from Decibel VC, Liquid 2, StoryHouse Ventures, Pioneer Fund, and Spot VC [The AI Insider, April 2026]. It was founded in 2025 by Georgia Witchel, a 24-year-old former professional ice climber whose path into deep-tech entrepreneurship has been documented in long-form interviews [Medium, March 2026]; [Startup Strides, 2026]. The technical premise is that disparate biomedical inputs can be combined into synthetic representations of anatomy, physiology, and behavior, addressing a persistent data-availability bottleneck in trial recruitment and biomedical device validation [TechCrunch, March 2026]; [Crunchbase, retrieved 2026]. The team remains small at three employees, consistent with a post-seed YC company in its first operational year [Y Combinator, retrieved 2026]. Investor interest sits at the intersection of two trends institutional capital has been tracking closely: the rise of in-silico trial methods (where Unlearn.AI is the most-cited reference point) and the broader application of generative methods to regulated, data-scarce verticals. Over the next 12 to 18 months, the watch items are first paying customers, the regulatory framing of any synthetic cohort outputs, and whether Mantis can convert YC and tier-one seed validation into a defensible data position before larger incumbents move into the same lane.
Data Accuracy: GREEN -- Confirmed across TechCrunch, Y Combinator, Crunchbase, and The AI Insider.
Taxonomy Snapshot
| Axis | Value |
|---|---|
| Stage | Seed |
| Business Model | B2B |
| Industry / Vertical | Healthtech (with sports and defense adjacencies) |
| Technology Type | Synthetic data, AI-driven digital twins |
| Geography | North America (San Francisco) |
| Growth Profile | Venture Scale |
| Founding Team | Solo Founder (Georgia Witchel) |
| Funding | Seed, ~$7.4M disclosed |
Company Overview
PUBLIC
Mantis Biotech was founded in 2025 by Georgia Witchel and incorporated as a San Francisco-based company through the Y Combinator program [Y Combinator, retrieved 2026]. The company describes itself publicly as building digital twins of humans, with a near-term emphasis on biomedical research and a longer-term framing that extends to sports performance and defense applications [Medium, March 2026]. Crunchbase additionally describes the company as building "a unified biomedical device testing and regulatory platform," a positioning that suggests the commercial entry point is closer to medical device validation than to drug discovery [Crunchbase, retrieved 2026].
The principal milestone to date is the company's seed financing, which was reported at approximately $7.4M and announced in early 2026 with Y Combinator listed as a participant alongside several institutional seed funds [The AI Insider, April 2026]; [Fyself News, retrieved 2026]; [Yahoo Tech, retrieved 2026]. Press coverage in late March 2026 by TechCrunch coincided with the funding disclosure and provided the first detailed third-party description of the product approach [TechCrunch, March 2026]. The company reports a headcount of three employees as of its YC profile, indicating that capital is still being deployed against early product and research work rather than a built-out commercial organization [Y Combinator, retrieved 2026].
Founder Georgia Witchel has spoken publicly about an unconventional path into the company, including a prior career in professional ice climbing involving high-risk expedition work [Startup Strides, 2026]. Authority Magazine's interview placed her at 24 years old at the time of the seed round and framed Mantis as the vehicle through which she intends to address "complex problems in medicine, sports, and defense" [Medium, March 2026].
Data Accuracy: GREEN -- Confirmed by Y Combinator, Crunchbase, TechCrunch, and Medium.
Product and Technology
MIXED
Mantis's stated product is a synthetic data engine that combines disparate biomedical inputs into digital twins representing human anatomy, physiology, and behavior [TechCrunch, March 2026]. The intended customers, per coverage and the Crunchbase description, are organizations that need access to richer biomedical cohorts than they can practically collect, with a near-term wedge in biomedical device testing and regulatory workflows [Crunchbase, retrieved 2026]. Mantis's own framing in interviews extends beyond healthcare to sports performance modeling and defense applications such as human-performance simulation, although no public customers have been named in either adjacency [Medium, March 2026].
The technical approach, as described in press, treats the digital twin not as a single high-fidelity simulation but as a synthetic dataset object that downstream researchers can sample from, condition on, or use to augment under-powered real-world cohorts [TechCrunch, March 2026]. This is consistent with how in-silico trial vendors have historically positioned: the deliverable is a statistically defensible synthetic control or augmentation layer rather than a clinical decision tool. The company has not publicly disclosed which model architectures, data partnerships, or validation methodologies underpin the system, and there is no public technical paper or benchmark to cite at this stage.
Founder interviews have flagged the ethical perimeter of predictive synthetic data as an active area of internal work, suggesting that governance and provenance of training inputs are part of the product conversation rather than purely a research concern [Medium, March 2026]. Hiring signal is limited: no open roles surfaced from the company's careers page or major ATS hosts at the time of research, so a tech-stack inference from job postings is not available here.
Data Accuracy: YELLOW -- Product framing confirmed by TechCrunch, Medium, and Crunchbase, but underlying architecture and validation evidence are not publicly documented.
Market Research and Opportunity
PUBLIC
The market backdrop for Mantis is a clinical research system whose largest single cost driver is delay, and where synthetic and in-silico methods are increasingly being treated as legitimate trial-design tools rather than research curiosities. According to coverage of Mantis's launch, roughly 80% of clinical trials are reported to experience delays attributable to data inaccuracies, with each delayed trial costing approximately $15M [Prism News, retrieved 2026]. While that figure is cited in launch press rather than in a primary sponsor study, it is directionally consistent with the broader pharmaceutical-industry literature on trial timelines and is the figure Mantis itself is using to frame the wedge.
| Sizing claim | Value | Source |
|---|---|---|
| Share of clinical trials delayed by data inaccuracies | ~80% | [Prism News, retrieved 2026] |
| Reported cost per delayed trial | ~$15M | [Prism News, retrieved 2026] |
Analyst takeaway: the cited numbers describe the size of the pain rather than the size of the addressable software market, but they explain why a synthetic-data wedge into trial design has attracted institutional seed capital. Even modest reductions in delay rates against a $15M-per-trial baseline produce enterprise-grade willingness to pay.
The demand drivers Mantis is positioned against are well established in the cited research: shrinking patient recruitment funnels for narrow indications, growing regulator openness to in-silico evidence in device submissions (the Crunchbase positioning around "biomedical device testing and regulatory" speaks directly to this), and the cost pressure on biopharma R&D budgets [Crunchbase, retrieved 2026]; [TechCrunch, March 2026]. The sports and defense extensions Mantis describes are adjacent rather than core; both are markets where human-performance modeling is an active procurement category, but neither has produced disclosed Mantis customers in public reporting [Medium, March 2026].
The most relevant regulatory force is the trajectory of FDA and EMA guidance on synthetic control arms and computational modeling in device submissions. Both agencies have published frameworks treating model-informed evidence as admissible in defined contexts, which is the policy tailwind that allows a 2025-vintage company to credibly target regulated workflows. The countervailing macro force is the same one any synthetic-data vendor in healthcare faces: payer, provider, and regulator scrutiny of training data provenance, bias, and reproducibility, which raises the evidentiary bar for any commercial deployment.
Data Accuracy: YELLOW -- Single-source sizing claim corroborated by category-level press but not by a named third-party market report.
Competitive Landscape
MIXED
Mantis is entering a category that already has a well-funded, late-stage reference player in Unlearn.AI, and its positioning will be read by investors against that benchmark.
| Company | Positioning | Stage / Funding | Notable Differentiator | Source | | --- | --- | --- | | Mantis Biotech | Synthetic datasets and digital twins for medicine, sports, defense | Seed, ~$7.4M | Cross-vertical framing (healthcare plus sports and defense); YC-backed | [TechCrunch, March 2026]; [Y Combinator, retrieved 2026] | | Unlearn.AI | Digital twins of trial participants for control-arm augmentation | Later-stage, multi-round venture-backed | Established regulator engagement and named pharma partnerships | [TechCrunch, March 2026] |
Analyst takeaway: the table understates how much of the in-silico trial conversation Unlearn.AI currently anchors. Mantis is not, on the public evidence, attempting to replicate Unlearn's pharma-control-arm wedge head-on; its Crunchbase framing around device testing and regulatory workflow suggests a different commercial entry point in the same broader category.
Segment by segment, the competitive map breaks into three groups. First, the dedicated in-silico trial vendors led by Unlearn.AI, which have spent years building the regulator relationships and statistical methodology required to have synthetic outputs accepted in pharma submissions. Second, large biomedical simulation incumbents (Dassault's Living Heart program, Sim&Cure, and similar device-focused simulation businesses) which already serve the medical device validation market that Crunchbase associates with Mantis. Third, a wave of generative-AI-for-biology startups, many YC-adjacent, that are applying foundation-model methods to molecular, imaging, or omics data and could plausibly extend into whole-body synthetic cohorts.
Mantis's defensible edges today are early and largely non-product. The clearest are capital and validation: a $7.4M seed with named institutional participants and a YC stamp gives it 18 to 24 months of runway and meaningful recruiting use in San Francisco [The AI Insider, April 2026]; [Y Combinator, retrieved 2026]. The cross-vertical framing (medicine plus sports plus defense) is also a positioning edge in fundraising conversations, because it allows a single technical platform to be sold against three distinct customer budgets. Whether that edge is durable depends on execution: cross-vertical narratives are perishable if the company does not produce a flagship customer in at least one segment within the next year.
The most exposed flanks are evidentiary and commercial. Unlearn.AI's specific advantage is a multi-year head start on the statistical and regulatory work that pharma sponsors require before they will accept synthetic cohorts in a submission, and Mantis cannot close that gap on capital alone. The medical device simulation incumbents own channels into device manufacturers that a three-person seed company does not yet have. The most plausible 18-month scenario splits along execution lines: Mantis becomes a winner if it lands a named device manufacturer or DoD-adjacent customer within the next 12 months and converts that into a reference deployment, validating the cross-vertical thesis. It becomes a loser if Unlearn.AI or a larger generative-bio entrant moves laterally into device testing before Mantis establishes a beachhead, in which case the seed round funds research rather than a defensible commercial position.
Data Accuracy: YELLOW -- Competitor identity confirmed by TechCrunch; relative stage and customer detail inferred from public reporting.
Opportunity
PUBLIC
If Mantis executes on its stated thesis, the prize is to become the default synthetic-data layer for human biology across regulated industries, a position with category-defining economics.
The headline opportunity. The largest plausible outcome for Mantis is to become the synthetic data infrastructure that biomedical device companies, pharmaceutical sponsors, and human-performance buyers default to when real-world cohorts are too small, too expensive, or too slow to assemble. The cited evidence makes this outcome reachable rather than purely aspirational: the same regulator posture that has allowed Unlearn.AI to embed in pharma trial design also opens the door for a device-and-defense focused entrant, and the reported $15M cost-per-delay figure establishes a willingness-to-pay envelope that supports enterprise contract sizes [Prism News, retrieved 2026]; [TechCrunch, March 2026]. A company that becomes the standard provider of synthetic cohorts in even one regulated submission category is a category-defining business.
Growth scenarios.
| Scenario | What happens | Catalyst | Why it's plausible |
|---|---|---|---|
| Become the default synthetic cohort provider for medical device submissions | Mantis wins reference deployments with two to three mid-cap device manufacturers and gets cited in successful FDA submissions | A named device-manufacturer partnership in 2026 or 2027 | Crunchbase positions Mantis specifically around device testing and regulatory workflow [Crunchbase, retrieved 2026] |
| Win a defense human-performance contract | A DoD-adjacent buyer procures Mantis digital twins for warfighter performance modeling | A SBIR or prime contractor partnership | Founder has publicly framed defense as a core vertical, not an afterthought [Medium, March 2026] |
| Cross-license synthetic cohorts to pharma | Mantis sells synthetic augmentation data to sponsors running underpowered Phase II trials | A named pharma pilot, comparable to Unlearn.AI's early partnerships | The reported 80% trial-delay rate gives sponsors direct economic motivation [Prism News, retrieved 2026] |
What compounding looks like. The flywheel for a synthetic-data company in regulated biology is data and reference deployments. Each accepted submission that incorporates Mantis-generated cohorts produces two compounding assets: a regulator-validated methodology that lowers the evidentiary bar for the next customer, and a richer training corpus that improves the next generation of digital twins. Distribution then compounds on top of methodology: device manufacturers and pharma sponsors talk to each other, and a single high-profile reference deployment typically pulls multiple follow-on customers from the same buyer network. There is no public evidence that this flywheel has started turning at Mantis yet, but the seed-round capital and team size are appropriate to the pre-flywheel stage.
The size of the win. The most useful comparable is Unlearn.AI itself, which has raised substantially more capital than Mantis on the strength of the in-silico trial thesis and is the reference asset investors will mark Mantis against. If Mantis reaches a similar position in device testing rather than pharma controls, a comparable later-stage valuation is a reasonable scenario, not a forecast. The broader prize is larger: regulated biomedical simulation, in-silico trial methods, and human-performance modeling collectively represent multi-billion-dollar enterprise budgets, and a company that establishes itself as the synthetic data primitive across more than one of those budgets is a candidate for category-defining outcomes.
Data Accuracy: YELLOW -- Scenarios grounded in cited press and category comparables; specific outcomes are explicitly framed as scenarios rather than forecasts.
Sources
PUBLIC
[TechCrunch, March 2026] Mantis Biotech is making 'digital twins' of humans to help solve medicine's data availability problem | https://techcrunch.com/2026/03/30/mantis-biotech-is-making-digital-twins-of-humans-to-help-solve-medicines-data-availability-problem/
[Medium, March 2026] Mantis Biotech CEO Georgia Witchel on Building Digital Twins, the Ethics of Predictive Data and Why Startup Founders "Are Not Special" | https://medium.com/authority-magazine/mantis-biotech-ceo-georgia-witchel-on-building-digital-twins-the-ethics-of-predictive-data-and-why-c14f46268b44
[Y Combinator, retrieved 2026] Mantis: Digital Twins of humans | https://www.ycombinator.com/companies/mantis
[Crunchbase, retrieved 2026] Mantis biotech - Crunchbase Company Profile & Funding | https://www.crunchbase.com/organization/mantis-biotech
[Startup Strides, 2026] Ep 58: From Ice Climbing to Startups: Georgia Witchel's Unconventional Founder Path | https://open.spotify.com/episode/3iRiRfIVczNg6nL1Yhqven
[Prism News, retrieved 2026] Mantis Biotech Builds Digital Twins of Humans to Solve Medicine's Data Problem | https://www.prismnews.com/news/mantis-biotech-builds-digital-twins-of-humans-to-solve
[Fyself News, retrieved 2026] Mantis Biotech is creating a digital twin of humans to help solve medical data availability issues | https://news.fyself.com/mantis-biotech-is-creating-a-digital-twin-of-humans-to-help-solve-medical-data-availability-issues/
[The Indian Practitioner, retrieved 2026] Mantis Biotech Uses AI-Powered Digital Twins to Transform Biomedical Research | https://theindianpractitioner.com/mantis-biotech-uses-ai-powered-digital-twins-to-transform-biomedical-research/
[Yahoo Tech, retrieved 2026] Mantis Biotech is making 'digital twins' of humans to help solve medicine's data availability problem | https://tech.yahoo.com/ai/meta-ai/articles/mantis-biotech-making-digital-twins-143000600.html
[The AI Insider, April 2026] Mantis Biotech Announces $7.4M in Funding to Advance AI-Driven Digital Twin Models for Biomedical Research | https://theaiinsider.tech/2026/04/02/mantis-biotech-announces-7-4-million-in-funding-to-advance-ai-driven-digital-twin-models-for-biomedical-research/
[Mantis Biotech, retrieved 2026] Mantis | https://mantisbiotech.com/
[LinkedIn, retrieved 2026] Mantis Biotech company page | https://www.linkedin.com/company/mantis-biotech
Articles about Mantis Biotech
- Mantis Biotech Wants a Synthetic Stand-In for Every Patient a Trial Cannot Recruit — The Y Combinator-backed startup is building digital twins of human anatomy and physiology, with $7.4M in seed funding and a tough regulatory road ahead.