Giving Scientists Their Time Back for Scientific Discovery
By Yunzhong, from Aofeisi
Scientists may spend far less time actually “doing research” than you might think.
When a new research topic arises, the starting point is rarely a flash of insight. More often, it is a long series of tedious preparations:
- Literature and data are scattered across different databases, spreadsheets, folders, and experiment records, requiring researchers to search through them one by one and painstakingly align the information;
- Key methods and experimental conditions are buried in the details of papers, and simply identifying and consolidating them can consume most of a researcher’s energy;
- After finally formulating a promising hypothesis, researchers still face environmental setup, parameter tuning, and repeated trial and error before they can validate it;
- Even after results are finally obtained, there is still a mountain of data to clean, charts to prepare, and conclusions to interpret…
In other words, the most valuable time scientists have—time that should be spent asking bolder questions, understanding deeper mechanisms, and making more meaningful discoveries—is often consumed by the “grunt work” of searching, organizing, validating, reproducing, and communicating.
So is there a way to give that time back to scientists?
Today, DP Technology has offered its own answer.
Its newly launched Bohr Scientific Space (Public Beta) is a desktop environment designed specifically for collaboration between researchers and AI scientists.
It connects downward to scientific data, computing, and experimental resources across local and cloud environments, while integrating AI scientists specialized in different disciplines directly into the research workflow.
Across the entire research process—from investigation, hypothesis formation, validation, and reproduction to figure generation and peer review—researchers only need to handle two things:
Asking questions and confirming results.
Scientific Space takes care of all the remaining “grunt work.”
Starting with Scientific Questions and Connecting the Entire Research Workflow
01 Investigation: More Than Information Gathering—High-Quality Judgment and Actionable Starting Hypotheses
When approaching a new research topic, the hardest part is not “finding more papers,” but distinguishing among the vast amount of information: What has become established consensus? What remains controversial? Which questions have yet to be answered?
In Bohr Scientific Space, researchers can pose research questions through dialogue.
SciMaster, the general-purpose research expert in Bohr Scientific Space, automatically searches literature and patents related to the research question, understands text, figures, molecules, and formulas in papers, and produces structured research findings.
Building on this, the system further organizes established consensus, contradictions, and gaps in the field, helping researchers formulate well-founded, testable research hypotheses.
When a hypothesis is worth pursuing, the AI scientist can also combine existing research, available data, and experimental conditions to generate an actionable experimental design, clearly defining the research variables, technical approach, key steps, and criteria for evaluating results.
Scientific investigation therefore moves beyond information gathering to become the starting point for high-quality judgment and subsequent research action.
02 Research Execution: Hypotheses No Longer Remain Discussions—They Can Be Validated with a Click
The truly time-consuming work often begins only after a hypothesis has been formulated.
Researchers must not only select appropriate models and methods, but also acquire data, install software, configure environments, set parameters, run tasks, troubleshoot failures, and analyze results. When something fails, they may need to try again with an entirely new set of parameters.
In Bohr Scientific Space, AI scientists can automatically break down tasks according to research objectives, invoke the relevant data, models, software, and computing resources, execute computations, simulations, and data analyses, and inspect results at key stages.
Researchers can continue making requests around the scientific question without repeatedly switching between different tools and working environments.
For fields including biology, medicine, drug discovery, and materials science, Bohr Scientific Space already includes AI experts such as BioMaster, PharmMaster, and MatMaster, supporting a wide range of real-world research tasks.
Researchers only need to ask the questions; AI scientists get the validation underway.
BioMaster: Making Biomedical Analysis Executable, Traceable, and Reusable
Bioinformatics research often requires constant switching among papers, public databases, code repositories, and analytical software.
The real time cost lies not only in running code, but also in finding the right data, reconstructing analytical methods, managing version dependencies, checking parameters, and determining whether the results have credible biological significance.
Consider three typical tasks:
Reproducing paper analyses: BioMaster can understand the methods described in papers, connect public datasets such as GEO with authors’ code, complete data acquisition, environment execution, code debugging, and result verification, and transform published papers into research assets that can actually be run and further utilized.
Single-cell data analysis: BioMaster can organize data processing, analytical workflows such as RNA velocity, and research visualization around a single scientific question, preserve earlier results, and allow researchers to append further analyses based on those results.
Interpreting genetic variants: For variants in genes such as TP53, BioMaster can align entities, verify evidence, conduct statistical analysis, and generate visualizations across data sources including ClinVar, UniProt, PDB, and AlphaFold, producing structured results with source information and evidence levels.
BioMaster organizes dispersed data queries and computational steps into an auditable analytical workflow, allowing researchers to devote more attention to evaluating results, understanding mechanisms, and designing the next stage of research.
PharmMaster: Connecting the Drug Discovery Workflow from Target Research to Molecular Design
Drug discovery involves multiple stages, including target research and project initiation, patent landscapes, mechanisms of action, structure–activity relationships, structural analysis, molecular design, and druggability assessment.
Each stage requires different information, models, and tools, while research conclusions must be continuously carried forward between upstream and downstream tasks.
Using the initiation and development of ITK inhibitors as an example, PharmMaster can:
- Gather extensive information around research objectives, accurately analyze patent and competitive landscapes, and accelerate project-initiation research and decision-making;
- Interpret SAR analyses through molecular recognition and contextual semantic associations;
- Automatically initiate computational tasks, complete starting-molecule screening, and carry out co-crystal structure analysis, retrosynthetic analysis, novel-molecule recommendation, and activity and ADMET property assessment, providing comprehensive intellectual support throughout the molecular discovery and optimization process.
Individual steps are no longer isolated tasks or complicated combinations of tools. Instead, they form a continuously advancing R&D workflow driven by the same research objective.
Researchers can inspect the rationale and results behind every step, adjust the design direction accordingly, and enter the next round of candidate-molecule optimization.
MatMaster: Organizing Materials Design and Simulation Around Performance Objectives
Materials research requires connections among literature evidence, structural design, performance prediction, and mechanism validation.
For example, in the development of high-temperature dielectric films, researchers need to systematically review relevant material systems, understand the relationship between structure and performance, and then use methods such as DFT and molecular dynamics to validate the stability and mechanisms of candidate materials.
Researchers can describe their target performance, operating temperature, and material constraints in Bohr Scientific Space.
MatMaster then conducts domain research, designs candidate systems, performs computational simulations, and analyzes the results, providing material directions worthy of further experimental validation along with the rationale behind them.
The focus therefore shifts from simply “finding a material that meets the requirements” to determining “which systems are worth validating first, and why.”
03 Research Outputs: Writing with Rigor and Reviewing for Credibility
Once the research is complete, one final hurdle remains:
How can the results be explained clearly? Are the citations reliable? Are the data and conclusions consistent? Can the evidence withstand reviewers’ questions?
In Bohr Scientific Space, the writing assistant can help researchers organize their thinking, structure papers, write rigorously, and produce content in formats such as LaTeX that can be further edited, based on research data and analytical results.
The review assistant verifies the paper’s logic, citation sources, figures and tables, and supporting materials, highlighting areas that require further confirmation or reinforcement.
Scientific成果 are built on the scientific knowledge accumulated across the world and throughout history. They must also re-enter the human knowledge system to undergo scrutiny, discussion, and validation.
Through writing and review workflows, Bohr Scientific Space helps researchers transform computational and experimental results into scientific成果 with clear structures, traceable evidence, and the ability to withstand continued professional scrutiny.
Why Can AI Execute Real Research Tasks in Bohr Scientific Space?
The key to bringing AI into the research process is not simply adding more model parameters or improving conversational capabilities. It is providing AI with a complete environment for understanding science, organizing tasks, invoking tools, and undergoing validation.
01 Scalable Research Resources: Connecting Knowledge, Computation, and Experimental Resources
Science Navigator connects to more than 200 million papers and patents, enabling AI to retrieve evidence from continuously updated scientific knowledge.
DeployMaster provides access to more than 50,000 carefully selected classic scientific computing tools. It works closely with DP Technology’s proprietary “DP YuZhi” models and tools, including large models such as DPA, Uni-Mol, and Uni-AIMS, as well as computing toolchains such as Uni-FEP and Uni-Dock, allowing models and software to be invoked instantly according to task requirements.
For wet-lab execution, Uni-Lab-OS can already connect to more than 150 categories and 1,800 types of instruments and equipment, and is continuing to connect with a broader experimental-instrument ecosystem. In the future, these data, samples, operations, and results will be integrated into a unified, closed-loop research workflow.
These rich, Agent-ready research resources provide the foundation for AI to execute real research tasks.
02 Multimodal Knowledge Understanding and Organization: Accurately Extracting Evidence and Systematically Mapping Knowledge Relationships
Uni-Parser can efficiently process multimodal scientific information, including papers, figures, molecular structures, formulas, tables, and experimental records.
For three types of structured content—text, mathematical formulas, and tables—its parsing accuracy exceeds 98% in each category.
At the same time, the system can effectively identify complex visual information such as figures, flowcharts, molecular formulas, and chemical reaction formulas, enabling dispersed data to be standardized, traceable, and converted into machine-readable formats.
Scientific Navigator and the Large Knowledge Model (LKM) further organize entity relationships, evidence sources, and research trajectories across literature, patents, data, and research records, enabling AI to understand established consensus, key disagreements, and research gaps surrounding a scientific question.
03 Agent Orchestration and Long-Horizon Execution: The SciX Framework with Strong Planning and High Fault Tolerance
The SciX agent framework breaks research objectives down into executable steps, coordinates knowledge, models, software, computing power, and experimental resources, and performs checks and validation at key stages. When a task encounters an obstacle, it can also diagnose and correct the problem automatically, giving it an exceptionally high level of task fault tolerance.
Sandbox provides batch, isolated, and recoverable execution environments for tasks, while supporting scheduling across heterogeneous computing resources. During a statistical window in which approximately 1.85 million launches were initiated, its creation success rate reached 98.8%.
AI can therefore move from a single research conversation into a complete research workflow, continuously carrying out investigation, hypothesis development, computational simulation, result analysis, and research-output generation.
04 Collaborative Evolution: Tools, Agents, and Humans Evolving Together Through Feedback from Real Tasks
Every search, reading task, computational invocation, and experiment execution forms part of the process for addressing the current question. Scientists’ corrections, computational results, experimental data, and lessons from failures collectively form a “scientific training ground,” providing long-term fuel for the accumulation and improvement of tools, agents, and human expertise, as well as valuable feedback for completing future research tasks more effectively.
As more workflows are executed, verified, and reused, validated methods can become reusable capabilities and workflows. Through the self-learning and self-evolution mechanisms of the EvoMaster agent, the platform will also “understand you better over time,” helping research organizations fully embed task experience into their broader operating systems while reducing repeated configuration and trial and error.
More updates on the technological advances behind the capabilities described above for Bohr Scientific Space will be shared with you in the near future.
For Researchers and Research Organizations
Bohr Scientific Space can serve both as a unified workbench through which researchers access knowledge, data, and tools, and as a collaborative environment where research teams organize AI scientists, reuse research processes, and accumulate research experience.
- For individual researchers, it reduces the time spent searching across databases, configuring environments, switching between tools, and organizing results, allowing the research process to be driven more by scientific questions than by tool operations.
- For research groups and R&D teams, it helps create traceable records of key methods, parameters, processes, and results, enabling validated analytical workflows to be inspected, reused, and iterated upon.
- For research institutions and corporate R&D organizations, it provides a unified environment for connecting internal data, computing resources, specialized tools, and domain expertise, supporting collaboration between human scientists and AI scientists on real-world tasks.
Bohr 107 Program: Finding the Scientific Questions Most Worth Solving in the AI Era
As research efficiency continues to improve, a more fundamental question is becoming increasingly important:
As more research processes can be assisted by AI, which scientific questions deserve to be explored first?
With this in mind, DP Technology has launched the “Bohr 107 Program,” inviting researchers worldwide to submit questions with significant scientific value and exploratory potential. Leveraging the capabilities of Bohr Scientific Space and its AI scientists, selected questions will receive research support spanning scientific investigation, hypothesis development, computational simulation, and result validation, helping valuable scientific questions progress from formulation to sustained exploration.
DP Technology hopes to work with researchers worldwide to continually identify important questions worthy of long-term investment and explore new pathways for human scientists and AI scientists to collaborate on scientific discovery.
Make important questions visible, give valuable ideas the opportunity to be validated, and enable AI scientists to help solve the scientific questions that are truly worth solving.
Giving Scientists Their Time Back for Scientific Discovery
AI scientists capable of continuously executing research tasks are developing rapidly. In the future, scientific discoveries will increasingly emerge from deep collaboration between human scientists and AI scientists.
But human scientists’ curiosity, judgment, and creativity will not be replaced by AI.
AI is better suited to handling the tedious, repetitive, yet indispensable work, giving scientists more time to ask more important questions, understand deeper mechanisms, and explore phenomena that have yet to be explained.
Let no piece of evidence be overlooked, let every idea have an opportunity to be validated, and let every exploration move closer to discovery.
Bohr Scientific Space is committed to building a research environment where human scientists and AI scientists can work together—an environment in which scientific research can continue to think, act, validate, and evolve.