OSS has developed a modular and extensible platform for AI Science Labs and Cloud Science Labs. OSS is layered platform that separates what scientists want to do from how experiments are executed in labs. It does this by prompting researchers to fully specify their experiments. In return, it provides details execution logs that ensure full auditability and traceability of every experiment step. By doing so, OSS bridges the lab across manual labs, automated labs, and AI-driven labs.
The OSS platform is built around three foundational ideas:
- Experiments must be fully specified: intent, parameters, constraints, and conditions need to be specified explicitly by researchers; this requires the platform to treat the experiment as “code” or “program”
- Execution must be deterministic and auditable: deterministic execution corresponds to reproducibility of science experiments; further, every step performed in lab should be tracked so that the experiment runs can audited, if needed
- Automation must be incremental: automation solution should be extensible so that the platform can support manual labs, hybrid setups, and fully automated labs.
This enables a smooth transition from traditional experimentation to cloud-based and AI-driven science, without forcing labs to be rebuilt from scratch.
The core of OSS is its Universal Lab Architecture, which consists of three tightly integrated layers:
- Universal Lab Abstraction (ULA) for experiment specification
- Universal Lab Engine (ULE) for orchestrating experiment execution
- Universal Lab Interfaces (ULI) for automating experiment execution
ULA allows researchers to define experiments in a device-independent format. ULE orchestrates and executes experiments precisely and repeatably. ULI connects the platform to real-world instruments and robots for automated execution.
Universal Lab Abstraction (ULA)
ULA is where experiments are defined in a device-independent, programmable form.
ULA allows researchers to focus on the scientific logic and experimental workflow, instead of the low-level details of lab instruments. To ensure that the experiment logic is fully specified, researchers are prompted to explicitly specify all parameters and conditions. ULA also supports control-flow constructs such as loops and conditionals so that researchers can express even the most complex experimental workflows easily.
To allow researchers to specify their programs, OSS supports logical lab actions such as transfer, mix, incubate, heat, measure-pH, measure-absorbance, and so on. As a result, researchers do not need to write instrument-specific scripts or vendor APIs.
This enables researchers to fully specify their experiments in a clean and precise manner. For example, experiment based on ELISA protocol shown here:
ULA, therefore, prompts researchers to describe what needs to happen scientifically because the OSS platform takes care of how it is executed by generating appropriate instrument instructions. This enables protocols (such as ELISA assays, serial dilutions, gradient mixing, etc.) and scientific experiments to be represented as reusable, executable workflows.
Universal Lab Engine (ULE)
ULE supports deterministic execution of the science programs by executing experiments exactly as specified. To do so, ULE supports various components that:
- Orchestrates steps in the correct order by managing timing and dependencies
- Provides robust execution by handling failures and retries
- Ensures safety constraints and preconditions are respected
- Captures complete execution logs and metadata
Execution of each instrument task and operator task generates a log, as shown. The execution records includes details about action, action parameters, experiment conditions, target instrument, and the outcomes (i.e., the raw results, metrics, and other signals generated during the task execution). These allow researchers to validate that the each step was executed as per their specification. These records can also be used for subsequent regulatory purposes.
ULE, therefore, ensures that every experiment is reproducible and auditable by design.
Universal Lab Interfaces (ULI)
ULI connects logical experiments to the physical world and the available instruments in the labs. Logical actions defined in ULA are mapped to a sequence of lab-specific, device-dependent tasks, which can be performed by either instruments or operators.
ULI supports two types of tasks: instrument tasks and operator tasks.
Instrument tasks: Mapped to vendor APIs that allow direct programmatic control. For example, OSS generates code to instruct an Opentrons liquid handler using its API, as shown.
Operator tasks: Defined as atomic and stateless tasks that human or robotic operators can perform when APIs are unavailable. For example, ULI maps “measure absorbance” logical action into a sequence of tasks that need to be performed using vendors’ digital interfaces, as shown below:
This design enables hybrid workflows – an operator task can be performed either by human operators or robotic operators. Indeed, human and robotic operators can execute tasks interchangeably or even collaboratively within the same experiment.
This allows OSS to work seamlessly across:
- Manual academic labs
- Shared national facilities
- Industrial R&D labs
- Partially automated or fully automated environments
As a result, OSS adapts to today’s labs, rather than requiring labs to adapt to OSS.
OSS Platform and AI Science Labs
The OSS platform provides the necessary and sufficient infrastructure for AI-driven experimentation.
- Fully specified experiments –> clear research intent and workflow
- Run-time records and rich task logs –> close-loop optimization across iterations and steps
- Deterministic execution –> foundation for safe and explainable AI interventions
- Instrument abstraction –> world model and control model reuse across labs
In short, the OSS platform turns physical experimentation into something AI can reason about, optimize, and improve – without sacrificing safety or reproducibility. As a result, the platform provides the necessary and sufficient infrastructure for AI-driven science, without requiring labs to be rebuilt from scratch.