Holonomy

Column-level encryption and masking for local Python analytics

Holonomy is a data minimisation tool that applies masking and hashing policies on personal data in local compute or containers. Single S3 can be shared across different roles for different processing purposes in Python analytics.

Read the Documentation →

pip install holonomy

How it works in 3 steps

1. Write Once with Encrypted Columns

Tag sensitive fields (pii, finance) in your schema contract. Holonomy writes standard Parquet files to S3 with sensitive columns locked via Parquet Modular Encryption (PME).

2. Define Access & Purpose in code

Map your corporate IdP groups (Okta, Azure AD) and business purposes (churn-model, marketing) to masking rules (HASH, REDACT, PLAINTEXT) in a cryptographically signed JSON policy.

3. Query Locally at Memory Speed

Analysts run holonomy.scan("s3://..."). The Rust core verifies their identity, unwraps permitted keys via your AWS/GCP KMS, transforms restricted columns in RAM, and hands a zero-copy Arrow stream to local engines.

Pragmatic Guardrails Instead of Enclaves

Most data leaks aren't malicious — they happen by accident when engineers pull production dumps into local notebooks. Holonomy ensures that authorized analysts and AI agents never accidentally pull raw PII into local memory or logs, helping with GDPR data minimization without slowing your team down.

Kill Defensive ETL Pipelines (One File, Many Views)

Stop duplicating S3 tables into users_raw, users_marketing, and users_anonymized just for masking purposes. Store one encrypted Parquet file on S3. Holonomy dynamically morphs columns at read-time based on who is asking and why.

Pseudonymization That Doesn't Break JOINs

Replacing emails or user IDs with NULL ruins cohort analysis. Holonomy’s in-memory deterministic SHA-256 hashing hides raw PII from data scientists while preserving primary-to-foreign key relationships across tables.

Zero Infrastructure to Maintain

No proxy servers, no Java clusters, no vendor lock-in. Holonomy runs 100% in-process inside your existing Jupyter notebooks, Airflow workers, or Kubernetes jobs with <5% memory-copy overhead. You keep full ownership of your S3 buckets and cloud KMS keys.

Early Adopter Licensing

Holonomy is currently in its final stages of development. The source code is made available under the Business Source License (BSL 1.1).

Although Holonomy requires a commercial license for production use, we are currently offering Early Adopter licenses completely free of charge, valid for full production use until June 1, 2027.

Reach out for early access keys and inquiries:
license@holonomy.eu

                    © 2026 Kirill Pruntov. All rights reserved.
                