Why Marigold?

The fundamental paradox of civic data management is that aggregation creates vulnerability. When a state agency centralizes 10 million voter records into a single AWS RDS instance to run deduplication scripts, they have inadvertently created a "honeypot" for state-sponsored threat actors.

Marigold solves this by pushing the compute down to the edge. We do not want your data. We want your data to stay exactly where it is. By utilizing WebAssembly and local-memory execution, Marigold shifts the burden of proof from a centralized server back to the local client. We distribute the vulnerability footprint to zero.

While Marigold is currently focused on securing and analyzing civic voter rolls, the underlying architecture is entirely dataset-agnostic. Marigold is fundamentally a DAIO (Data Analysis and Inference Optimizer).

Our core value proposition is simplifying complex statistical analysis for non-technical users on massive, local datasets. Whether the data is a state voter roll, highly regulated medical billing records (HIPAA), or complex financial ledgers, the Marigold DAIO engine can map, traverse, and execute probabilistic inferences on the data entirely in local memory.

To support the DAIO paradigm, Marigold utilizes a Cartridge architecture. A Cartridge is a modular, community-built correlation matrix that runs natively within the Marigold local-compute environment.

Researchers can use our LLM assistant (Mari) to build custom Cartridges. For example, an academic researcher might ask, "What is the statistical correlation between LED streetlights and violent crime incidents?" Mari generates the execution logic into a deployable Cartridge. The user executes the Cartridge against their local dataset, and Marigold returns plain-English data visualizations and Z-Score confidence intervals—all without exfiltrating the raw data.