DP Technology (深势科技) has launched Bohrium Science Space in public beta, a desktop research environme...
The AMW Read
An established AI-for-science player packages multiple named domain agents and an orchestration/sandbox framework into one autonomous research platform, meaningfully expanding its product baseline within the AI agents segment without resolving a cross-segment debate.
DP Technology (深势科技) has launched Bohrium Science Space in public beta, a desktop research environment that embeds AI "scientist" agents directly into the scientific workflow rather than offering a single prediction model.
The platform bundles several named agents: SciMaster for general literature review and hypothesis generation, BioMaster for bioinformatics analysis and paper reproduction, PharmMaster for target research through molecule design in drug discovery, and MatMaster for materials design and simulation. These are backed by Science Navigator, which indexes over 200 million papers and patents, and DeployMaster, which deploys more than 50,000 scientific computing tools alongside DP Technology's own models (DPA, Uni-Mol, Uni-AIMS) and tools (Uni-FEP, Uni-Dock). Uni-Lab-OS extends the system into physical experiments, connecting over 150 instrument categories and 1,800 devices. An orchestration layer called SciX breaks research goals into executable steps and checks results at each stage, reporting a 98.8% run-success rate across roughly 1.85 million sandboxed task launches; writing and review assistants then help draft and check manuscripts. DP Technology also opened "Bohrium 107," a call for global researchers to submit high-value scientific questions for platform-backed investigation.
The launch shows an AI-for-science strategy shifting from point models toward autonomous, multi-step agents that plan, execute, and validate research end to end, tying literature retrieval, compute orchestration, and lab-instrument control into one system — a bet that proprietary data and tool integration, not just a stronger base model, is what makes domain agents useful in practice.
For builders and investors evaluating similar vertical-science-agent plays, the open question is whether DP Technology's wet-lab automation and hypothesis-generation claims hold up under independent replication, since no third-party benchmark accompanies this launch; per the AI Market Watch index, DP Technology is linked to 172 patent records (patent linkage covers roughly 800 of the ~5,000 companies tracked), a signal worth watching for whether this platform converts into licensing or external adoption beyond internal tooling.
#DPTechnology #AIforScience #AIAgents #DrugDiscovery #MaterialsScience #AIInfrastructure
