Palantir Foundry — Platform Overview

What is Foundry?

Palantir Foundry is an enterprise data integration and analysis platform that provides a unified operating system for data. It enables organizations to integrate, transform, analyze, and act on data at scale. Foundry sits at the core of Palantir's commercial offerings and is used by governments, defense agencies, healthcare systems, and Fortune 500 companies.

Core Architecture

Foundry is built on three foundational pillars:

1. Data Integration Layer

The data layer handles ingestion from hundreds of source types via Magritte connectors. Data lands in Foundry as Datasets — immutable, versioned snapshots with branch-based development (similar to Git).

• Raw Datasets: Ingested directly from connectors
• Derived Datasets: Produced by transforms (Python, PySpark, SQL)
• Branches: Isolated development environments — merge changes with confidence

2. Transformation & Logic Layer

Code Repositories house Python and PySpark transform logic. The platform handles compute scheduling, dependency resolution, and incremental computation automatically.

# Example: Python Transform
from transforms.api import transform, Input, Output

@transform(
    output=Output('/path/to/output/dataset'),
    source=Input('/path/to/source/dataset'),
)
def compute(output, source):
    df = source.dataframe()
    result = df.groupBy('entity_id').agg({'value': 'sum'})
    output.write_dataframe(result)

3. Ontology Layer

The Ontology translates raw data into business objects — Patients, Orders, Missions, Equipment. Applications built in Foundry query the Ontology rather than raw datasets, creating a clean separation of concerns.

Key Products Within Foundry

Foundry Security Model

Foundry implements column-level, row-level, and dataset-level access controls through Markings. Sensitive data can be restricted to specific groups, roles, or clearance levels. All data operations are audited.

Deployment Models

• Cloud-hosted SaaS: Palantir-managed on AWS/Azure/GCP
• Customer-hosted: Deployed in customer VPC via Apollo
• Air-gapped / JWICS: Disconnected environments via Apollo (common in DoD)

Getting Started

1. Request access to your Foundry enrollment URL
2. Navigate to the Data Catalog to browse available datasets
3. Open Code Repositories to write your first transform
4. Use Contour to explore transformed data visually
5. Define Ontology objects in the Ontology Manager