01
XtalPi Science:
Uniting Agentic AI, Scientific AI, and Physical AI for autonomous science discovery
Human Scientists
Access
Invoke
capabilities
capabilities
API
Model access
External Models
Human Scientists
Science Token
Invoke models
API
Model access
External Models
02
Agentic AI
Multi-Agent: an intelligent hub that understands scientific intent and orchestrates discovery end to end.
We use Agentic AI to connect human scientists, domain models, and Physical AI labs into one autonomous workflow.
Agentic AI is transforming how scientific work gets done. Instead of isolated tools and handoffs, intelligent agents interpret research goals, plan experiments, and coordinate digital prediction with physical execution—so teams move faster from question to verified result.
At the center is Multi-Agent system—our intelligent hub that orchestrates Scientific AI models and Physical AI robotics, while connecting to external foundation models when needed. The result is a closed loop of reasoning, action, and learning across the discovery lifecycle.
Core principles
01
Intent
understanding
Agents translate scientific objectives into actionable plans—prioritizing tasks, constraints, and success criteria across discovery scenarios.
02
Cross-system
orchestration
Agentic AI coordinates domain models, robotics, data pipelines, and external models into a unified workflow scientists can supervise and steer.
03
Closed-loop
autonomy
From planning to execution to analysis, agents continuously iterate—feeding experimental outcomes back into the next cycle of discovery.
Accurate
Accurate across various targets with the state-of-the-art molecular force field
Comprehensive
Covers drug design scenarios like binding free energy, scaffold hopping, and protein mutations.
Virtual Screening
Ultra-large chemical libraries require efficient screening. Active learning enhances virtual screening by iteratively training AI models.
03
Scientific AI
Proprietary AI Models: Our industry-leading models help scientists discover & optimize better molecules, faster
We harness Scientific AI and physics-based methods to explore chemical space accurately & efficiently
Scientific AI is advancing molecular discovery by generating diverse candidate compounds and exploring vast chemical space. High-throughput virtual screening, enhanced by active learning, helps identify promising candidates more efficiently.
Physics-based free energy perturbation (FEP) predicts binding affinities at accuracy comparable to experiment across a wide range of design scenarios. Combined with state-of-the-art molecular force fields and cloud-scale computing, Scientific AI becomes a practical engine for modern R&D.
Core principles
01
Generative
molecular design
AI compound generation unlocks ultra-large chemical space and covers the majority of molecular design scenarios and generation strategies.
02
High-throughput
screening & design
Highly efficient algorithms and mass allocation of cloud compute deliver screening and design at industrial scale.
03
Accurate binding
affinity predictions
Physics-based free energy perturbation predicts protein–ligand binding at accuracy comparable to experiment—across diverse drug discovery scenarios.
03
Physical AI
Lab Robotics: Transforming laboratories with Physical AI for higher efficiency, precision & throughput.
We deploy embodied lab robotics and digitized systems to execute chemistry at scale—and return high-quality experimental data.
In modern scientific research, Physical AI is reshaping the wet lab. Automated systems make high-quality, high-volume experimental data more accessible than ever—raising efficiency and freeing researchers from repetitive manual work.
We build intelligent laboratories that combine robotics, automation, and digital infrastructure so AI can act in the physical world—improving precision and throughput in chemical execution and data generation, and powering the next wave of data-driven discovery.
Core principles
01
Automation of
experimental processes
Advanced automated systems streamline experimental workflows across chemistry and materials workflows. Scalable automation increases efficiency and consistency of research operations.
02
Digitization of
laboratory data
Digital systems help researchers capture, access, and analyze experimental data with higher accuracy and reliability—accelerating problem-solving in the lab.
03
Intelligent laboratory
infrastructure
Built on automation and digitization, our Physical AI infrastructure enables autonomous experimental execution and analysis—raising research quality so scientists can focus on higher-order experimental design and interpretation
Accelerate autonomous science discovery
Use Agentic AI, Scientific AI, and Physical AI to discover new molecules—faster, with greater rigor, and grounded in physical experiment.