Laeka / SentinelX
Code language detection. For those who protect.
An adaptive linguistic detection API for anti-trafficking organizations. The piece that's missing from existing detection stacks. Free for nonprofits.
The linguistic brain existing tools don't have.
Trafficking codes mutate every 4-8 weeks. Keyword lists are always behind. SentinelX detects patterns of dissimulation in text — not keywords, behaviours. It adapts in days, not months.
Built on the OmniQ protocol: detection lenses that read patterns of perception, not words. Multi-agent architecture proven on Starpod. A nonprofit that understands nonprofits — peer, not vendor.
"It's not a competitor to Spotlight, Arachnid, or Traffic Jam. It's the missing layer — the intelligence that plugs into what you already run."
Risk score. Category. Signals. Reasoning.
The reasoning is the killer feature, not the score. It cuts analyst review time by reducing why-is-this-flagged investigations.
Codes mutate. We adapt.
Two-stage pipeline: lexicon + embeddings filter ~90% of noise (≤200ms), then Claude structured output classifies risk + signals + reasoning. Verified cases feed back into few-shot examples — the model learns new codes in days.
Canadian hosting. Zero US cloud.
Data sovereignty is a deal-breaker for the organizations we serve. Local-first, encrypted, HIPAA/PIPEDA-aligned. Your data, your jurisdiction.
Reasoning, not just scores.
Every flag comes with the LLM's explanation. Investigators spend 70% less time figuring out why a text was flagged. Faster triage, fewer false positives, more victims found.
A 90-day pilot is the deal.
Side-by-side with your existing detection. Target: 30%+ reduction in analyst review time. Free for verified nonprofits. Reach out — we'll listen first, then propose.