C12.AI — CONNECTING BITS & ATOMS

Physical AI Scientists

Connecting molecular design with physical execution in real laboratories, helping scientists move faster from hypothesis to validation.

Watch Demo Contact Us

Trusted By Industry Leaders

BeOne Medicines

Risun Group

Qilu Pharmaceutical

Research institutes under the Chinese Academy of Sciences

Our Mission

Helping Good Science Happen Faster

We see too much valuable time in laboratories still consumed by repetitive execution, condition adjustment, and cross-instrument work. We want to hand more of that burden to reliable systems, so scientists can focus on judgment, creativity, and the questions that matter most. If validation becomes faster and lower-friction, more valuable molecules, medicines, and materials have a chance to reach reality sooner.

Why Now

In molecular science, the bottleneck is often not design, but validation

AI can now generate candidate molecules and experimental ideas in a short amount of time, but physical execution in the real world remains slow, expensive, and difficult to scale. That stretches the loop from hypothesis to validation and limits the steady accumulation of high-quality experimental data. Design capabilities are advancing quickly, while physical execution remains a key constraint on R&D speed.

>$10B

Per approved drug

~10yr

Development timeline

<10%

Success rate

Digital AI

Minutes / molecule

Physical execution

Weeks / molecule

We bridge this gap with Physical AI Scientists.

Core Capabilities

Closing the Loop from Design to Execution

We are starting with some of the most real and demanding parts of drug R&D, then extending these capabilities to broader molecular science workflows.

Molecular AI Brain

Integrating retrosynthesis planning, molecular property prediction, spectral analysis, and the interpretation, analysis, and hypothesis adjustment of experimental phenomena, it helps teams move experimental loops forward more efficiently and shortens the time from molecular design to validated results.

Physical AI Execution

Powered by our proprietary VLA architecture, our general-purpose mobile robot works directly with existing lab instruments, taking on repetitive, cross-instrument, and adjustment-heavy execution work while reducing manual switching and wait time.

Data Flywheel

Every real experiment generates new execution and feedback data that continuously improves the system, allowing our robot to build a deeper understanding of the physical world over time.

Closed-loop Workflow

AI Design

Physical Execution

Auto Analysis

Continuous Evolution

Loop

Starting from drug discovery, our robot is equally applicable to R&D and manufacturing across the broader molecular sciences.

See our robot in Action

We are starting with some of the most real and demanding parts of drug R&D, then extending these capabilities to broader molecular science workflows.

Validated Results

Proven inside a top global pharma lab

Real-world molecule purification success rate

0 days

Trial operation at BeOne Medicines' lab

Physical operations with zero execution errors

BeOne Medicines · Small-Molecule Purification (Project Code: Talos)

Deployed inside BeOne Medicines' production-grade pharmaceutical lab. Our robot autonomously executed drug molecule intermediate purification with real-time chromatographic condition adjustment and cross-instrument operation. Zero lab renovation. Full human-robot collaboration.

Purification of 15 real drug intermediate molecules
Stable operation in a real R&D environment
Zero lab renovation, plug-and-play deployment
Collaborative execution of complex workflows with scientists

Core Team

A team grown at the intersection of healthcare and AI

We come from the intersection of pharma, healthcare, and technology, so we understand not only models and systems, but also the workflows, constraints, and collaboration patterns inside real laboratories.