Not a wrapper. A patented engine.
EQIQs runs on a three-layer stack: a proprietary scoring engine, a provider-agnostic LLM layer, and a proprietary data flywheel. The math is patented. The language is generative. The corpus is ours.
Proprietary Scoring Engine
Twenty-one established psychometric instruments synthesized into one context-weighted compatibility score. Deterministic, auditable, and HR-defensible — the math is protected by U.S. Patent #9,002,961.
- 21 frameworks: Big Five, MBTI, DISC, Enneagram, Attachment, EQ, RIASEC, Values, and more
- 23-dimension compatibility scoring across Romantic, Business, Co-parenting, Friendship
- Role-adjusted weighting and dampened proxy formulas for defensibility
- Every score reproducible, explainable, and traceable to the underlying instruments
LLM Orchestration Layer
A provider-agnostic AI gateway routes to frontier models (Gemini 2.5, GPT-5) with automatic fallback, prompt-injection guards, and per-request cost logging. LLMs generate the language; the patented engine generates the number.
- Multi-provider gateway — no vendor lock-in, models swappable in one line
- Automatic fallback chain when a provider degrades or rate-limits
- 15+ production AI surfaces: coaching, onboarding, clarifier, /profile/analyze, /compatibility/explain
- Prompt-injection detection and full transcript logging for compliance
Proprietary Data Corpus
Every quiz response, framework result, and compatibility outcome is stored and normed. Over time this becomes the training set that lets us fine-tune small models on tasks no competitor can replicate — because they don't have the data.
- Structured response corpus growing with every assessment
- Longitudinal outcomes data linked to real relationship and hiring contexts
- Foundation for future fine-tuned models on 'given these scores, predict X'
- Cannot be scraped, cannot be reproduced without our user base
"Are you building your own LLM?"
Short answer: no — and that's deliberate. Training a foundation model from scratch is a nine-figure capex problem with no defensibility at the end. What we're building is more valuable:
AI Shipping Today
The orchestration layer powers real product surfaces — not a demo, not a roadmap slide.
Why This Compounds
Four independent moats that reinforce each other — the reason EQIQs is priced as AI-native infrastructure, not another SaaS wrapper.
Patent Moat
U.S. Patent #9,002,961 protects the context-weighted synthesis method. Copycats cannot legally claim the same category.
Framework Moat
21 established instruments with normed scoring. Reproducing this takes years and licensing agreements — not weeks.
Provider-Agnostic Moat
Not dependent on any single AI vendor. When the next model wins, we adopt it in one deploy.
Data Moat
Every user session grows a corpus that cannot be scraped. The longer we run, the harder we are to catch.
Where We Are & What's Next
Multi-provider gateway, 15+ AI surfaces, prompt safety, cost logging — all in production.
Eval harness for every AI surface. Structured output enforcement. Reasoning-model support on high-stakes calls.
First fine-tuned small model on our corpus for compatibility explanation — 10x cheaper, faster, and more on-brand than general-purpose LLMs.
Full retrieval-augmented layer over longitudinal outcomes. Predictive fit models for hiring and team formation.
Built on secure, compliant infrastructure
Row-level security, TLS 1.3, AES-256 at rest, GDPR-ready controls, and full audit logging on every AI call. The stack is protected by U.S. Patent #9,002,961.