Palantir

Metropolis

Integrate, enrich, model, and analyze any kind of quantitative data.

The Platform

On the back-end, the Palantir Metropolis platform comprises a suite of capabilities for integrating tabular data from many different sources into coherent models, performing complex computations over models, and sharing and iterating on analytic products.

Data Integration

The Palantir Metropolis platform fuses data from multiple sources into a unified representation called a model. The platform includes a library of pre-built adapters, frameworks, and APIs that allow it to easily integrate data from disparate source systems for analysis.

Metrics

Metrics are the heart of computation in the Palantir Metropolis platform. Metrics composite, aggregate, transform, compare, and perform calculations over models. Palantir Metropolis ships with a built-in library of widely-used metrics, including change and percentage change, moving averages, derivatives, integrals, and various other fundamental time series and mathematical manipulations.

Custom Metric Development

Hedgehog Language, or HHLang, is a scripting language with syntax similar to Java that was specifically developed to facilitate complex analysis in Palantir Metropolis—models, metrics, and documents are all first-class language constructs in HHLang. With language features such as a proper type system, expression chaining, anonymous functions / lambdas, and collections, HHLang allows analysts to describe both simple expressions and complex, multi-module calculations.

Complex custom metrics can be written in a built-in integrated development environment (IDE) that supports code completion, interactive debugging, and modular code design practices. HHLang comes with an extensive library for quantitative analysis that covers time series analysis, regression, statistics, and advanced date/time handling.

Data Modeling

In Palantir Metropolis, data models are the basic building blocks of analysis. Models are the translation of the rows and columns of source datasets, including descriptive metadata, into a unified conceptual object that represents an entity in the world. A model can be an organization, a company, a person—any real world object described by the data. Each installation of Palantir Metropolis is configured with the types of models necessary to answer the questions at hand.

Iteration and Collaboration

Analytic products created within the Palantir Metropolis front-end applications are stored as units of sharable, linkable, recombinant analysis called documents. Documents are not static, finished products, but living works-in-progress that can be iterated upon and included as inputs into new analysis. Users can construct complex documents by chaining together many simpler documents, then share and collaborate on them with other users across the enterprise.

Extensibility, Customizability, and APIs

Palantir Metropolis is designed from the ground up to be extensible at every layer of the stack. From low-level data integration, custom metrics, to building custom user interface to implement specific workflows, it has been designed as a fundamentally open platform.