Platform Architecture

The 3-Space Vector Model replacing static segmentation.

The Core Architecture: Three Interlocking Vector Spaces

We move beyond simple segmentation into predictive intent modeling by architecting three distinct "Vector Spaces" that interact:

1. Semantic Content Space

(The "What")

We train models to vectorize every URL in our 3 billion behavioral logs.

  • Reverse-Engineering Search Intent: We interpret the likely query that led to a visit.
  • Conceptual matching: Understanding "Security" without needing the keyword.

2. Hierarchical Identity Space

(The "Who")

Most companies treat users as flat lists. We map Business Team Hierarchies.

  • Graph Embeddings: Mapping relationships (Individual -> UPID -> Business Team).
  • Inference: "Junior Developer" reading docs implies "VP of Engineering" is buying.

3. Angle & Messaging Space

(The "Pitch")

Your Angle Library turned into math.

  • Vectorized Library: Subject lines and hooks are vectors.
  • Distinct Angles: "Save Money" vs. "Beat Competitors" are distinct mathematical directions.

The 1:1 Personalization Engine

How the AI performs a mathematical search to find the perfect message:

1

The "Session" Vector (Real-time Intent)

A user visits a URL. The system looks at their last 5 actions and combines them into a Current Intent Vector.
Interpretation: "User is researching 'implementation speed'."

2

The "Identity" Context (The Filter)

The system checks the Identity Graph Embedding for this UPID.
Interpretation: "This user is a CTO at a Series B Tech company."

3

The Match (Vector Search)

The AI searches the Messaging Library for the vector closest to:
Intent ("Fast implementation") + Role ("Strategic oversight").

4

The Extrapolation

Result: The email isn't generic. It says: "Stop struggling with latency. Here is how [Product] helps teams like [Company Name] implement faster..."