MGI Tech Co., Ltd., a Chinese sequencing instrument manufacturer, has established an artificial intelligence subsidiary named Genoria AI to develop and commercialize laboratory tools that can autonomously plan and execute experiments, aiming to streamline biomedical research workflows.
Dr. Yang Meng, MGI Tech's Chief AI Officer and CEO of Genoria AI, explained in a recent interview with the international gene industry media GenomeWeb that the subsidiary was created to align with the global trend of integrating AI with biology. "We want to send a clear signal to the public and the market that we take AI business very seriously," said Yang, who also noted that the CAIO role was established just last month at MGI Tech.
Founded in March, Genoria AI is headquartered in Shenzhen, like its parent company. It focuses on AI4S, specifically building a closed-loop infrastructure for life sciences that integrates AI agents, agent-ready automated experimental systems, and real-world experimental feedback loops. "We are trying to combine these two technologies to build autonomous laboratories that can scale the next wave of scientific discovery," Yang stated.
The AI model developed by Genoria AI in collaboration with the Shanghai Artificial Intelligence Laboratory is called ProtoPilot. According to Yang, ProtoPilot is designed as a closed-loop AI agent that can start from a scientific goal, formulate an experimental plan, and then convert that plan into executable code for instruments, enabling automated workflows. After execution, ProtoPilot learns from the experimental results to guide subsequent optimization. The model can operate autonomously or under human supervision.
In a validation study published on the arXiv preprint server earlier this month, Yang's team demonstrated ProtoPilot's capabilities across 294 tasks and four increasingly complex wet-lab workflows, including real-world applications in synthetic biology and molecular biology, such as plasmid construction and mutant engineering. However, ProtoPilot is still under development. Yang noted that Genoria AI is currently focusing on training the system to understand experiments by collecting human expert thought processes to train the underlying large language model. The desired outcome is that researchers can describe their experimental goals in natural language, and ProtoPilot will convert them into experimental plans and then generate machine-executable scripts.
Genoria AI and MGI Tech aim to deploy ProtoPilot across multiple fields, including sequencing library preparation. MGI Tech previously had a similar initiative, collaborating with Chulalongkorn University researchers in Thailand to develop an AI agent called PrimeGen for designing targeted sequencing primers, published in Nature Biomedical Engineering last year. Beyond sequencing, Yang said Genoria AI wants to apply ProtoPilot to other areas like enzyme and protein engineering, as well as single-cell and spatial omics.
Currently, Genoria AI's business model involves selling AI lab software and hardware products to customers. "Our current positioning is as a solution company, helping our clients or collaborators make their labs smarter and more autonomous," Yang said. The company's primary target customers are large national laboratories, followed by AI-focused biotech companies, pharmaceutical firms, and contract development and manufacturing organizations (CDMOs). Existing collaborators include the Fujian Lujiang Innovation Laboratory and the Shanghai Artificial Intelligence Laboratory.
Some of Genoria AI's products are "AI-native" automated instruments. Yang added that MGI Tech's existing lab automation platforms, such as the PrepALL liquid handler, can be upgraded with AI computing power to become lab agents. Regarding pricing, Yang stated it varies by project, with some entire lab-level projects potentially costing millions of dollars. In the long term, Genoria AI also aims to use AI agents to co-develop new intellectual property with customers and collaborators, such as designing and manufacturing enzymes or other molecules in autonomous workflows. Yang noted that the company may initially focus on selling AI-driven automation platforms, and its instruments are compatible with any AI model chosen by researchers.
MGI Tech and Genoria AI are not alone in pursuing the opportunity of AI autonomous labs, a rapidly emerging concept. For instance, Ginkgo Bioworks is also developing its own autonomous AI lab products, integrating software and hardware to help design and conduct experiments. AI companies like Anthropic and Phylo have launched AI workbenches called Claude Science and Biomni, respectively, to assist scientists in biomedical research and analysis. Yang emphasized that compared to some peers, Genoria AI benefits from its parent company's decade-long expertise in multi-omics tools and hardware manufacturing. "We are not a single model company. One of our major advantages is that we start from the manufacturing and lab side," she said.
Whether and when AI agents can reliably help researchers scale their discoveries remains to be seen. Yang noted that currently, most lab AI agents cannot truly convert scientific ideas into experiments and discoveries without human intervention. "They may only solve one part of your entire experiment, but they still cannot complete an end-to-end workflow. And even if AI agents can act autonomously, they will not replace human experts. They may replace standardized and repetitive work, but great ideas will still come from humans, or from the collaboration between humans and AI."