Radical Numerics
Radical Numerics is an AI research laboratory developing general biological intelligence through large-scale, multimodal AI models tailored for biological data. The organization advances architectures, numerical methods, and hardware-efficient systems for scientific AI, enabling progress in therapeutic discovery, genome analysis, biodefense, and scalable biological intelligence.
Industries
N/A
Nr. of Employees
small (1-50)
Products
A scalable language model trained on genomic sequences for applications in human health and biodefense, enabling variant interpretation, genome engineering, and pathogen detection.
A scalable language model trained on genomic sequences for applications in human health and biodefense, enabling variant interpretation, genome engineering, and pathogen detection.
Expertise Areas
- AI-based genomics and sequence modeling
- Multimodal and long-context AI architecture development
- Numerical methods for scientific machine learning
- Custom systems engineering for AI training
Key Technologies
- Genome language models
- Multimodal large-scale AI architectures
- Blockwise quantization protocols (FP4, FP8)
- Stochastic rounding for low-precision training
Key People
News & Updates
Raised $50M in seed capital to build scalable AI for biological applications and global resilience.
Announcement of a $50 million seed round to develop general biological intelligence and a preview of a next-generation genome language model for health and defense applications.
Preview of initial biodefense results using genome language models for functional pathogen detection and fitness scoring of viral variants beyond standard alignment-based methods.
Research preview of a next-generation genome language model applied to generative genomics and variant analysis.
Technical analysis of custom kernels, dataflow, and system optimizations enabling efficient FP4 training in large foundational AI models.
Discussion of floating point fundamentals, mixed-precision training, and stability techniques for large-scale AI models using FP4.
Raised $50M in seed capital to build scalable AI for biological applications and global resilience.
Announcement of a $50 million seed round to develop general biological intelligence and a preview of a next-generation genome language model for health and defense applications.
Preview of initial biodefense results using genome language models for functional pathogen detection and fitness scoring of viral variants beyond standard alignment-based methods.
Research preview of a next-generation genome language model applied to generative genomics and variant analysis.
Technical analysis of custom kernels, dataflow, and system optimizations enabling efficient FP4 training in large foundational AI models.
Discussion of floating point fundamentals, mixed-precision training, and stability techniques for large-scale AI models using FP4.