UViiVE

Autonomous science from bits to bioactives

Explore the Platform
Scroll

Why Now

Three things converged in 2026.

🤖

AI Agents That Do Science

Not just assistance — specialized agent fleets, shared memory, 24/7 autonomous execution. With trust.

⚙️

Physical Robots Closing the Loop

Molecular design to Kd measurement, functional experiments, and beyond. No human required.

🧬

De Novo Peptide Design

This unlocks topical bioactives at scale, and MoCRA is the regulatory arbitrage for rapid commercialization.

We build the agentic infrastructure that does science end to end — and we point it first at AI-designed precision peptides for skin.

About Us

The scientific staff
is an agentic fleet.

UViiVE is a research company built the other way around. The scientific staff is a persistent fleet of AI agents — each with its own identity, its own tools, and a shared memory that survives the session that produced it. Our scientists set direction and hold the gate on what leaves the fleet.

That fleet is the company. It designs experiments, mines the literature, writes and runs code, ranks candidates, and carries what it learns forward into work it has never seen before. We published what happened when we ran it continuously for six months, including the trust architecture we had to build before a shared memory was safe to recurse on.

Skin peptides are our 1A. Transdermal delivery of AI-designed precision peptides is the first problem we pointed the fleet at — a real program with a real loop, chosen because it closes fast enough to teach the system. It is the leading example of what the fleet does, not the boundary of it.

10
Persistent AI agents
24/7
Autonomous execution
104
Multi-phase reasoning cycles
Everything we claim about the fleet is in the preprints below, with the methods, the benchmarks, and the limitations.

Technology

The AI-Native Full Stack

Ten agents. Shared memory. 24/7 autonomous execution.

Literature & hypothesis

Searches and retrieves the literature, mines prior work, generates hypotheses, and stores null results so they are searchable before new work begins.

Design & simulation

Structural biology, protein folding, and binder design on local GPU hardware. Turns candidate lists into ranked hits overnight.

Code & analysis

Writes and executes code, runs analyses, spawns sub-agents, and reports results with full provenance back into shared memory.

Autonomous robotics

Liquid handling, cell-free synthesis, and functional assays. The atomic layer that closes the loop while we sleep.

The Exponential Unlock

A Full Stack Agentic Fleet

The AI fleet designs. Robots execute: from in silico candidate to atomic experiment.

A design-test-rank loop that runs 24/7, gets smarter with every cycle.

Research

We publish the machinery.

Every claim below is drawn from the preprint it sits under. Follow the DOI for the full method, data, and limitations.

Full Stack Agents · FleetbioRxiv, August 2026

A Persistent Fleet of AI Scientists Exhibits Cooperative and Autopoietic Behavior

  • A fleet of cooperative AI scientist agents ran continuously for nearly six months on shared memory, tools, and cross-agent communication
  • An identity-level fabrication constraint cut delusion-reinforcement probe failures from 91.7% to 0%
  • Fleet-specific project memory lifted a local open-weight model on internal benchmarks from 44% to ~90%
  • 104 recurring multi-phase reasoning cycles produced 43 manually curated hypotheses
doi:10.64898/2026.08.16.745122
VaaSmedRxiv, March 2026

VaaS is a Multi-Layer Hallucination Reduction Pipeline for AI-Assisted Science

  • 225 annotated rare-disease gene curations plus a prospective 100-gene collection, together over 3,000 verified citations
  • Unguided LLM output produced 95.9% wrong-topic (Type II) citation hallucination; the full protocol reached 0.0% Type I and 6.5% Type II, a >14-fold reduction
  • Live PMID verification alone eliminated both error types entirely (0.0% / 0.0%)
  • F1 = 0.9853 on the MedHallu hard tier, against a published GPT-4o baseline of F1 = 0.811, at under ~$1 per comprehensive gene review
doi:10.64898/2026.03.24.26348935
MIRRORmedRxiv, May 2026

Combined values alignment and epistemic verification prevent delusional reinforcement in conversational AI agents

  • 156 scored responses across 39 probe items in four belief-harm domains, under triple-blind human consensus rating
  • The bare language model passed only 3 of 36 main-battery probes (8.3%)
  • All three safety architectures passed at least 97% of items (P < 0.001 versus the bare model)
  • On the synergy battery, only the complete architecture passed every item; single-layer conditions failed 7 of 16 (43.8%)
doi:10.64898/2026.05.29.26354389

Team

Meet the Team

Deep expertise in AI, protein engineering, and clinical dermatology — united to revolutionize precision biology.

SE
Stephen C. Ekker, PhD
Co-Founder & CEO
  • ·Discovered VEGF-A inhibition stops UV-induced erythema
  • ·Serial entrepreneur
  • ·Professor and Dean, UT Austin
MP
Milit Patel
Co-Founder & CTO
  • ·AI/ML scientist
  • ·Biotech, stats & data science
  • ·Protein engineer
WW
Wesley A. Wierson, PhD
Scientific Founder
  • ·Serial entrepreneur
  • ·Y Combinator backed founder
  • ·Translational protein engineering
  • ·Gene editing & cell therapies
DR
Drew Rice
Head of Product
  • ·Former Chief of Staff, Capital Factory
  • ·Quantitative developer & data expert
  • ·Founder; Austin startup ecosystem
SH
Stella Hartono, MD/PhD
Scientific Founder
  • ·Discovered VEGF-A inhibition stops UV-induced erythema
  • ·Clinical assistant professor, Stanford
  • ·Immunologist & allergist
VB
Victoria Bedell, MD/PhD
Scientific Founder
  • ·Discovered VEGF-A inhibition stops UV-induced erythema
  • ·Anesthesiology & critical care

And the Fleet

Ten persistent agents, each with a standing specialty and a shared memory the others can read.

Uvy
Fleet operations

Coordinates programs across the fleet and maintains its shared tooling.

Atlas
Research infrastructure

Architecture review, provenance and paper audits.

Zevo
Genetics

Genome editing and model-organism genetics; coordinates science across the fleet.

Buster
Protein engineering

Binder design, scoring and design validation.

Kodiak
Protein design

Validation and analysis of designed proteins.

Mendel
Rare disease genetics

Curates the rare disease database and supports genetics programs.

Lyra
Verification

Checks claims and citations before they leave the fleet.

Galen
Review and synthesis

Cross-project review and synthesis.

Helios
Inference

Serves inference workloads for the rest of the fleet.

Pip
Benchmarking

Rehearsal and canary testing before work reaches the fleet.

Vision

Closed loop. Autonomous.

Design→
Build→
Test
Bioactives

Get in Touch

Interested in learning more? Drop us a message.