Thank you! What we have seen done in the past is use things like pybind11 to expose C++ APIs in Python, which I guess is a similar concept to SWIG. If you are

Thank you! What we have seen done in the past is use things like pybind11 to expose C++ APIs in Python, which I guess is a similar concept to SWIG. If you are using build tools such as Bazel, you can even get the C++ compiled at run-time when submitting your pipeline. Regarding i/o artifacts, Sematic lets user choose how they are serialized. We offer reasonable baseline defaults, but certain artifacts require serialization formats that are cross-language (e.g. ROS messages since you mention robotics). As a last resort, users are free to serialize and persist artifacts by hand as part of their pipeline functions (e.g. storing in a cloud bucket) and only returning a reference to said artifact (e.g. a Python dataclass with artifact location and metadata).