Bota Launches SAION AI – "Physical AI" Platform for Biomanufacturing
San Francisco and Hangzhou, China – As artificial intelligence continues to transform the digital world through cognitive and generative capabilities, a new breakthrough has emerged: Physical AI – an intelligent system capable of understanding, reasoning, and acting in the physical world.
Today, Bota launched SAION AI, the first Physical AI platform for biomanufacturing.
SAION AI focuses not merely on in silico design (computer simulation-based). The platform represents a comprehensive Physical AI suite that integrates cognitive, orchestration, and execution capabilities in a single system through unified experiments. With this approach, SAION AI can continuously optimise biological discovery and biomanufacturing processes.
SAION AI is designed with a three-layer architecture: Cognition, Orchestration, and Execution.
Supported by large language models (LLM), the platform unites scientific reasoning processes with real-world experimental execution. This architecture streamlines coordination from understanding biological systems through laboratory experiments, creating a closed-loop cycle that continuously learns and optimises biomanufacturing processes.
Cognition: Biological Understanding at Multiple Scales
The Cognition layer is built on data from Bota’s Cell2Cloud Biofoundry, integrating tens of millions of experimental data points, millions of scientific publications and patents, and various public biological databases. Combined with AI4Science models, SAION AI can develop biological understanding across multiple scales, from genes, proteins, and cells to fermentation processes. This approach facilitates more systematic design of biological systems and data-driven scientific decision-making.
Orchestration: Intelligent Research Coordination
The Orchestration layer centres on an Agent orchestration engine that leverages the reasoning capabilities of LLMs to coordinate multi-agent collaboration, utilisation of various scientific tools, and overall research workflows.
This layer can decompose complex research objectives into structured tasks and integrate 316 specialised scientific instruments, supporting automated research workflows with fault tolerance capabilities.
Execution: Automated Experiments
Through Bota’s Biological Protocol Language, SAION AI can convert experimental designs into standard instructions directly executed by laboratory equipment. Data from actual experiments is then sent back to the Cognition layer. In this manner, AI models can be continuously updated, improving research and development (R&D) performance.
Real-World Performance
SAION AI has demonstrated state-of-the-art (SOTA) performance across various AI benchmarks in the life sciences, proving its capabilities as an AI Scientist system.
Key achievements include:
Scientific literature understanding: 70.7% for LitQA + SuppQA, surpassing several leading general-purpose AI models.
Biological sequence reasoning: 88.2% for SeqQA when working on DNA, RNA, and protein tasks.
Genetic engineering: 84.9% for gene editing and cloning benchmarks.
Scientific discovery: 89.6% for BAIS-SD benchmarks.
Real-world validation demonstrates that SAION AI autonomously completes entire research processes—from literature review through laboratory experiment assembly—with accuracy exceeding 90%.
Towards Autonomous Biomanufacturing
The launch of SAION AI represents a crucial step towards a new era in biomanufacturing. Processes previously dependent on trial-and-error methods now evolve into intelligent engineering disciplines, specifically as AI and physical laboratories work together to accelerate scientific discovery and industrial-scale production.