Mantis

Physics-Enhanced Datasets for Low-Resource Machine Learning

Winter 2026ActiveB2BInfrastructureArtificial IntelligenceSports TechBiotechBiotechnologyNew York, NY, USA

Company

Mantis is a digital twin company that combines LLMs with high-fidelity physics simulations in order to convert rare and difficult-to-access human behavior data into predictive models.

Verdict

High Signal

Market Opportunity

Clinical trial data quality is a massive and well-documented problem in life sciences. 80% of clinical trials face delays from data inaccuracies; average $15M lost per trial. The global clinical data management market is $1B+ and growing. ICP is clear: pharma companies, CROs, and biotech teams managing clinical datasets. Enterprise B2B with obvious ROI (preventing $15M data-related trial losses) makes the monetization path straightforward.

High Signal

Founder Signal

Georgia Witchel: Harvey Mudd CS (elite engineering school), UW Masters Bioengineering, CTO of Louiza Labs (autonomous surgical robotics, raised $5M), Algorithm Engineer at Theo Health (built and commercially launched pose estimation software, attracted Olympic gold medalist Xander Schauffele as investor), Tesla intern (11 months). Elite ice climber (4 World Cups, 8 world records, first to represent climbing at Olympics in Lillehammer 2016) — exceptional domain depth in human performance data. Claims to have worked with HPE software as an athlete, innovated from an ML context, and built commercial solutions on top of it — uniquely credentialed intersection. Repeat funded founder in biomedical space.

Medium Signal

Competition

Life sciences data platforms have strong incumbents: Medidata (Dassault Systèmes), Veeva Systems (Vault CDMS), Oracle Clinical One, and Parexel all compete in clinical data management. The 'biological meaning encoded into canonical reusable datasets' angle is a differentiated positioning vs. raw data management tools, but it needs to prove durability against Medidata's distribution advantages. Not in the synthetic data space (previous competitor note was wrong) — this is clinical trial data quality, where switching costs are high once embedded.

Medium Signal

Product

Raised $6.3M seed from YC + Pioneer Fund (a respected biotech-focused VC) and has a growing team of 4-5 people (Software Engineer, AI Fellow @ UCLA, Biostats PhD @ UNC, Business & Ops). Product is described specifically: domain-aware data platform for life sciences that encodes biological/clinical meaning into reusable datasets for clinical trial data quality. No named clinical trial customers publicly, no pricing page, no demo — but the funded team and Pioneer Fund backing suggest the thesis has been validated with at least some potential customers.

OverallA Tier

Mantis is more substantial than the empty website suggests. Georgia Witchel is a genuinely rare founder: Harvey Mudd CS + UW Bioengineering + CTO of a $5M-funded surgical robotics company + commercial pose estimation product + elite athlete who understands human performance data from every angle. The $6.3M seed from Pioneer Fund (biotech-specialist VC) signals real due diligence and at least some customer validation. The clinical trial data quality market is large with clear ROI. Risks: solo founder in a complex enterprise sales cycle, no named customers public yet, and Medidata/Veeva have entrenched distribution in pharma. But the founder-market fit here is exceptional and the funding validates the thesis enough to move this to A.

Active Founders

Georgia Witchel

Founder

I'm an Ex-elite ice climber turned repeat computer science / biomedical founder. After bringing climbing to an Olympic stage (Lilihammer 2016), breaking 8 world records, and developing the sport of ice climbing, I dropped out of high school, got a degree in CS from Harvey Mudd, was a founding engineer for a sports tech company, founded a YC company (Louiza Labs), and earned a master's in Bioengineering.

Mantis

TierA Tier

BatchWinter 2026

Team Size3

StatusActive

LocationNew York, NY, USA

Last Updated3 days ago