04

Trust & safety.

Moderation as a recommendation problem.

<2s
decision turnaround
0.4%
false-positive rate
What we build

Moderation as a recommendation problem. Productionized.

Real-time classifiers for image, text, and audio combined with human review queues, appeals workflows, and a transparency layer. Built for DSA, COPPA, and regional regimes from day one rather than retrofitted under pressure.

Common deployments

Image and profile moderationBehavior-based abuse detectionRegulatory reporting

What we ship

  • Real-time classifier pipeline with monitoring.
  • Reviewer console with documented SOPs.
  • Appeals workflow and transparency report templates.
  • DSA-ready audit trail and quarterly export pipeline.
System overview

Five components, one cohesive runtime.

We build each layer to operate independently, then compose them into the brands we run for our members.

01layer 1 of 5

Real-time classifiers

Image, text, audio, and behavior models scoring every action.

02layer 2 of 5

Decision orchestrator

Risk-weighted action: allow, soft-block, queue for review, ban.

03layer 3 of 5

Human review console

Reviewer tooling with case context, history, and consistent UX.

04layer 4 of 5

Appeals workflow

User-facing dispute path with regulator-ready audit trail.

05layer 5 of 5

Transparency reporting

Quarterly metrics export ready for DSA disclosure.

Tech we reach for

We pick boring, proven tools where boring works. Where the problem demands more, we build deliberately on top of these primitives.

PyTorchONNX RuntimeTritonPostgresTemporal
Outcomes

What it looks like in production.

<2s

Time-to-action on flagged content

0.4%

False positive rate (target)

100%

DSA reporting coverage

Get in touch

Building the systems that bring people together.

Questions about our brands, press, or partnerships? Send a note and we'll reply within two business days.