Why 15 Years of Ontology Experience Shows Palantir’s Ontology Isn’t New

2011‑2015: Early Exposure – IEC 61850, OPC UA, and Wonderware ThingWorx

The author began work during China’s digital substation boom, encountering heterogeneous vendor equipment and the need for interoperability. IEC‑61850 provided a standardized, semantically rich protocol, replacing ad‑hoc address‑to‑meaning mappings with uniform field definitions, supported by a national technical specification.

Around 2014, the rise of Industry 4.0 prompted exploration of OPC UA, whose strength lies not only in communication but in its information model—a highly abstract language of nodes defining objects, variables, and methods that can interoperate across standards.

2016‑2018: Platform Practice – Custom Object Models and Device Operations

Joining a startup that built a low‑code platform based on object models, the author designed a diagnostic model for Baowu Group’s equipment. The model remained at code and database‑table level without a formal modeling language, and the development process was reactive (“head‑pain‑to‑head‑pain”) without a top‑down ontology methodology.

2018‑2020: Platform Product – OPC UA‑Based Ontology and Visualization

After multiple projects, the author advocated for model‑driven platforms, noting Europe’s push for semantic standards. The planned software expressed both device data and the platform itself via information models, likening the whole system to Palantir’s ontology. Challenges included the high complexity of modeling for low‑end customers, leading to heavy simplifications, and the need for a repository of reusable models.

2020‑2024: AI Practice – Small and Large Model Projects

Working at Baidu, the author engaged in industrial AI projects, including knowledge graphs. While small models were common, large models were emerging. The focus shifted to unstructured data governance and integrating knowledge graphs with large language models, using a tiered logic for non‑structured data.

2025‑Present: Re‑launch – Palantir Narrative and Machine‑Readable Standards

With the popularity of large models and Palantir, the author seeks to narrate AI data using DIKW, three‑layer ontologies, information models, knowledge graphs, and unstructured data governance. The missing bridge among these concepts was addressed by adopting a machine‑readable framework to connect them.

Conclusion

Concepts such as semantic, kinetic, and dynamic ontologies have existed long before Palantir’s offerings. While Palantir emphasizes dynamic ontologies, earlier work focused on semantic and kinetic aspects, highlighting differing emphases across the industry.