INFevo Technology
INFevo builds a closed loop from data and AI Virtual Cell models to automated experiments, while developing an autonomous discovery system in which hypotheses are traceable, evaluable, and falsifiable.
Platform definition
Beijing Infinite Evolution Technology Co., Ltd. (INFevo) develops AI Virtual Cell (AIVC) foundation models that learn and predict transcriptional changes under genetic, chemical, and environmental perturbations. The focus is not static classification, but the relationship between perturbation, response, and state transition.
The platform connects model predictions to automated cell experiments. Computation produces testable judgments, experiments return real feedback, and the results inform the next model and hypothesis cycle. This loop supports research into disease mechanisms, candidate targets, and drug screening.
Core capabilities
OCOO-T is a virtual-cell model for transcriptional perturbation-response prediction. The public preprint uses flow matching and a standard Transformer to model continuous gene expression while encoding perturbation type, dose, and cellular context.
Research focuses include generalization to unseen perturbations, cell types, and experimental conditions, and unified modeling across genetic, chemical, and cytokine perturbations.
The automated cell experimentation platform organizes culture, perturbation, readout, and quality control into reproducible workflows. Model outputs can enter real experiments, and experimental results return to computation in structured form.
The loop is designed to shorten the cycle from question to prediction, validation, and revision while preserving experimental boundaries, uncertainty, and negative results.
The Popper Project (TPP) is INFevo’s autonomous scientific discovery system currently in internal testing. It organizes evidence retrieval, mechanism reasoning, hypothesis generation, experimental design, and feedback.
The system emphasizes traceable processes, evaluable hypotheses, and falsifiable conclusions. Its role is not to replace scientific judgment, but to make evidence chains, decisions, and next experiments clearer.
Discovery loop
Specify the biological question, testable hypothesis, and evaluation criteria.
Combine multi-omic data, virtual-cell models, and existing evidence.
Test key judgments in reproducible cell experiments and record uncertainty.
Return results to models, knowledge, and hypotheses to guide the next cycle.
Research directions
Predict cell-state responses to gene edits, compounds, doses, and environmental changes across cellular contexts.
Form testable hypotheses around disease-associated states, regulatory networks, and candidate mechanisms to support target prioritization.
Use models to narrow the experimental search space, then validate candidate interventions and response direction through automated experiments.
Public evidence
Model methodology, experimental setup, and public benchmark results.
Benchmark results and internal testing status for the discovery system.
Collaboration around AI Virtual Cells and organoid validation.
Join us
We are hiring across algorithms, biology, cell experimentation, and automation.