Drug Safety Intelligence

Billion dollars to learn the drug didn't work?
We tell you first.

Nuclear-grade risk engineering applied to drug toxicity. Every prediction traceable. Every number auditable.

See live demo How it works

What If You Could Have Known?

VIOXX — MERCK, 1999–2004

$4.85B

in settlements

88,000

excess cardiac events (est.)

In early validation, our model flags Vioxx at 25mg as disease-risk — consistent with the APPROVe trial that found 2x heart attack risk after 5 years and thousands of patients.

12.5mg

BORDERLINE

25mg

DISEASE

50mg

DISEASE

Preliminary model output. Full validation in progress.

Per-drug cost rises exponentially through clinical phases — we flag failure at preclinical stage

The Problem

90% of drugs fail in clinical trials. $2.6B wasted per drug.

Nuclear reactors are stress-tested with rigorous fault tree analysis. Every failure path is mapped, measured, and audited.

Drug development? Pick a target. Run trials for a decade. Hope it works. Nobody maps the biological failure logic — until Phase III burns half a billion dollars.

90%

Clinical trial failure rate

$2.6B

Cost per approved drug

Average cost from discovery to FDA approval. Most of it wasted on drugs that fail.

10+ years

Discovery to market approval

What We Deliver

Drug risk answers before the first patient is dosed.

Four deliverables. Each one traceable to its source.

Risk quantification

A single risk score per compound at any dose. Not a heatmap — a number with traceable logic behind it.

Driver identification

Which biological pathways drive risk? Ranked by quantitative impact, not literature frequency.

Dose-response curves

Continuous risk profiles across dose ranges. See exactly where safety thresholds are crossed.

Full audit trail

Every node traces to PubMed. Every number has a reason. FDA-ready documentation from day one.

$500M+

SAVED PER KILLED PROGRAM

Kill failing drugs before Phase III burns half a billion dollars

Why

NOT JUST WHETHER

Know which biological pathway drives risk — not a black-box guess

Days

NOT A DECADE

Risk profile before IND filing — not after 10 years of trials

100%

AUDITABLE

Every claim traced to PubMed. Walk into FDA with confidence

Proven Methodology

Nuclear PRA, applied to drug development.

Probabilistic Risk Assessment has kept nuclear reactors safe for 60 years. We applied the same framework to biology.

Vital Risk Assessment (VRA) Overview

Three layers: biomedical knowledge base → biological failure paths → quantitative risk profile

Adapted from Probabilistic Risk Assessment for the International Space Station, Smith (2002)

Traditional AI

Pattern matching

Correlation-based predictions
Black-box neural networks
Millions of uninterpretable edges
No causal reasoning
Cannot explain why a drug fails

VRM (Vital Risk Model)

Causal engineering

Fault tree logic (AND/OR gates)
Every node traceable to literature
Quantitative risk propagation
Auditable by domain experts
Shows exactly why risk increases

Why Us

The only platform with causal, auditable drug risk.

Others find correlations. We map mechanisms.

VRM
Causal +
Auditable +
Quantitative

Correlation-based AI — Outer ring

Drug repurposing via statistical patterns. No causal model, no audit trail, no risk numbers.

Structured AI — Middle ring

AI target identification with some structure. Still a black box. Not quantitative.

VRM — Core

Full causal fault tree logic + quantitative risk scores + every node traceable to literature.

USPTO Patent Filed (2026) — Sponsored by UIUC Office of Technology Management

Google DeepMind × Drive Capital Hackathon (2026) — Winner (LLM-driven pipeline)

UIUC National Center for Supercomputing Applications (NCSA) LLM Hackathon (2025) — Winner

Defensibility

Triple moat.

Three compounding barriers to entry.

FDA

ISTAND

PATENT

TREE

LIBRARY

VRM

Patent

USPTO patent filed for applying PRA to drug safety. First-mover protection on the core methodology.

FILED — USPTO

Tree Library

Every validated fault tree is a proprietary asset. Each takes months of synthesis. The library compounds over time.

GROWING ASSET

FDA ISTAND

Pursuing qualification through FDA ISTAND. Once qualified, VRM becomes the regulatory gold standard.

IN PROGRESS

Market Timing

Nice-to-have today. Must-have after FDA qualification.

FDA ISTAND transforms VRM from competitive advantage into regulatory requirement.

Now

ISTAND Filing

Post-Qualification

NOW

Nice-to-have

Pharma uses VRM to de-risk early pipeline decisions. Competitive advantage for forward-thinking teams.

ISTAND FILING

Growing adoption

FDA engagement signals regulatory direction. Early adopters build internal expertise.

POST-QUALIFICATION

Must-have

FDA-qualified methodology becomes the standard. Every drug safety submission references VRM-class analysis.

Team

Built by engineers and scientists.

Jaemin Yang

Founder

UIUC — Nuclear Eng. (BS, MS), Informatics (PhD candidate)

George Vazquez

Co-founder

Merck

Leah Kim

Designer

Fermilab. Ex-Boeing, ex-Microsoft.

Catherine Blake

Advisor

Professor, UIUC. Carle Illinois College of Medicine (CIMed).

Jodi Flaws

Advisor

Professor, UIUC. Cancer Center at Illinois.

Join us.

We're looking for

BD Lead — Customer discovery, pilot contracts, fundraising
PRA PhD — Fault tree design, ISTAND docs, case studies
LLM PhD — Pipeline optimization, extraction accuracy, compound expansion

GET IN TOUCH

Billion dollars to learn the drug didn't work?We tell you first.

Billion dollars to learn the drug didn't work?
We tell you first.