CADMUSAI

Contextual intelligence engine powering Mundial Media

Mundial Media Inc. April 2026

Engine Overview

Tri-modal contextual intelligence for multicultural ad targeting. Built by Ramón Cendejas.

3
Analysis Modes
(Text + Image + Sentiment)
750+
In-Culture Publisher Sites
3x
Brand Awareness Lift
vs. Non-Contextual
4
Unsafe Content Clusters
(Crime, War, Disasters, Sensitive)

Core Purpose

Classifies publisher web pages in real-time to deliver contextual ads without cookies. Uses only the content the reader is currently viewing, not user identity or browsing history.

Cultural Differentiator

Uses domain-native training data curated by multicultural engineers instead of keyword blocklists or standard IAB categories. Understands Spanglish, code-switching, and bilingual patterns natively.

Why It Matters

General-market tools over-block culturally relevant content: immigration reporting flagged as "controversial," Dia de los Muertos imagery as "morbid," "slays" in Black cultural context as "violence." Cadmus knows the difference.

How Ads Get Served

Contextual segment IDs injected into ad request before reaching Google Ad Manager (GAM). Pre-delivery brand safety, not post-bid verification where waste has already occurred.

Key distinction: Cadmus is backend-only, no user-facing front end. It powers the contextual analysis feeding into Mundial's other platforms (Aries, Publisher Platform, Consumer Insights) and campaign delivery through GAM.

Processing Pipeline

Crawl to ad delivery in five stages

Stage 1

Ingestion

Dual-Path:
Offline (Batch): Continuous crawlers parse publisher pages, extracting URLs, text, images. Trains and updates models.

Real-Time (Active): When a user visits a page, a Dynamic Ad Tag fires. Active Parser extracts HTML, splits into text + images.

Stage 2

Tri-Modal Analysis

Three analyzers run in parallel:

NLP / Text Semantic meaning, topics, cultural context, bilingual patterns
Image / Deep Learning Objects, scenes, visual context, explicit content detection
Sentiment Emotional tone, crime reporting vs. glorification

Stage 3

Brand Safety Scoring

Weighted formula using proprietary omega weights tuned on multicultural data. Same content on different publisher types triggers different calculations.

Fast-Fail: Explicit content or blacklisted segments = immediate rejection.

Stage 4

Classification & Injection

If page passes thresholds, Cadmus assigns Contextual Segment IDs injected into ad request before reaching GAM. GAM matches IDs against campaign targeting. Ad serves only if segments match.

Stage 5

Caching & Persistence

In-Memory Cache: Fast lookups on repeat ad requests for the same page. Articles are analyzed once, results cached.

Database: Persistent storage for classification history and model training feedback loops.

Brand Safety Clusters

Four top-level unsafe content clusters:

1. Crime

Dense clustering, clear boundaries. High confidence in detection and separation from safe content.

2. War & Conflicts

Dense clustering, clear boundaries. Distinguishes reporting on conflict from glorifying it.

3. Disasters

Moderately clustered. Requires contextual judgment to separate news coverage from exploitative content.

4. Sensitive Topics

Broadly distributed. Requires nuanced weight tuning. This is where cultural calibration matters most.

Evolution

Word-level clustering to tri-modal contextual engine

2022
V1

Word-Level Clustering + Basic Imaging

Custom word embeddings from a domain-specific corpus (not a general LLM). Unsupervised clustering to find natural groupings (athletes, musicians, etc.) without forced categories. Human labeling step to name clusters. Basic image analysis via AWS Rekognition. Already generating revenue at this stage.

Word Embeddings Unsupervised Clustering Basic Image (AWS Rekognition)
2023
V2

Sentence-Level Analysis + Sentiment + Advanced Imaging

Moved from word-level to sentence-level analysis. Added sentiment analysis (reporting on crime vs. glorifying it). More advanced image analysis. System now understands meaning in context, not just word presence.

Sentence-Level NLP Sentiment Analysis Advanced Imaging
2025
V3 - Current

Enhanced Sentiment + Brand Safety Overhaul + Basic Video

Brand safety overhaul: word variants (k1ll, ki11, s3x) mapped back to canonical forms so they cluster correctly instead of being smeared across all clusters. Dramatically improved safety scoring accuracy. Basic video support via metadata extraction (not frame-by-frame). Tri-modal parallel processing (text + image + sentiment).

Brand Safety Overhaul Enhanced Sentiment Basic Video (Metadata) Variant Normalization
TBD
V4+

To Be Defined

Ramón should define the next major version with input from those who talk to brands and publishers. Current wishlist items (see Roadmap) are incremental. The challenge is identifying the next "big jump."

Needs Definition

Infrastructure

Shared AWS, separate schemas per platform

Mundial Media Platform Architecture

CADMUS AI

Backend only. No front end.
Contextual engine.

ARIES

Internal platform.
Backend + Frontend + DB

PUBLISHER PLATFORM

Publisher-facing.
Frontend + DB

CONSUMER INSIGHTS

Newest. Prod only.
No dev environment yet.

↓ All on shared AWS ↓
AWS (Primary)

Same database, different schemas per platform

Google Ad Manager

Segment IDs injected before ad serving

Environment Status

Platform DB / Schemas Dev Environment Prod Environment Notes
Aries (Internal) Manages data for what the Publisher Platform shows
Cadmus AI Backend-only engine. No front end.
Publisher Platform Frontend-only. Data fed from Aries backend.
Consumer Insights ? Javier spinning up dev env. Currently prod-only. Reads from same prod DB.
Sandbox Environment (Proposed): Sandbox for non-engineers (vibe coders) to safely connect to copies of data without risking production. Ramón agrees this is needed. Not yet created.
Ukrainian Contractors: Own local machines, WordPress/HTML migration. Not on AWS. Potential to absorb future lower-complexity tasks.

Roadmap

Prioritization TBD with leadership

Cadmus Engine Improvements

High

Dynamic Classifiers from Historical Data + Keyword Search

Build classifiers on the fly from past content for ad hoc insights. Sales team's top ask: gives them a reason to contact clients proactively. Achievable relatively quickly per Ramón.

Medium

Scroll Depth Weighting (Above the Fold)

Weight content above the fold higher than below, especially on mobile. Untested but expected to improve accuracy.

Medium

Upgrade IAB Categories to Google's Taxonomy

Google's taxonomy has more topics than IAB. Confirmed after reviewing competitor (Grana) approach. Expands segment granularity.

Infra

Process Improvement: Python to Go

Migrate extraction pipeline from Python to Go for speed. Helps with dynamic classifiers and faster document iteration. Rust considered but Go has better library support.

Medium

Enhanced Video Descriptions

Use LLMs to generate richer video descriptions from titles and metadata. Previously blocked by Google API costs, now feasible with Claude.

Explore

Full Video Analysis (a la GumGum)

Frame-by-frame video analysis (transcribe audio, capture one frame/sec, run image analysis). What Grana/GumGum does. "Completely different animal" per Ramón. Would not build on top of Cadmus.

Adjacent Product Ideas

High

First-Party Demographic Data Matching

Integrate first-party demographic data with Cadmus contextual data. Currently using third-party sources (SimilarWeb, Semrush). Would resolve the tension between "we don't need cookies" and needing demographic info.

Medium

Supply/Demand Content Guidance for Publishers

Tell publishers what topics will have upcoming ad demand so they can create matching content. E.g., "Automotive campaigns increasing 15% next quarter." Ramón's idea. No one currently offers this.

Marketing

Cadmus Step-by-Step (Behind the Scenes)

Demo material showing how Cadmus processes a page step by step. Sales enablement, not a product feature.

Cadmus Health Baseline (Proposed)

Proposed: Establish measurable baselines across three dimensions. Currently no systematic way to track improvement, compare versions, or benchmark against competitors.

Uptime

Availability and downtime tracking.

Speed

Page processing latency. Critical for real-time ad serving.

Relevancy

Classification accuracy. A/B testing across versions. Competitive benchmarking.

Team

Cadmus and broader tech ecosystem

Core Team

RC

Ramón Cendejas

CTO / Cadmus Creator
Sole architect of Cadmus across all three versions. SOW mandates 70%+ of his time on Cadmus.
JV

Javier

VP of Engineering
Supports Cadmus. Leads publisher development. Authored the Technical Reference Card.
AB

Alejandra Brizuela

Head of Product Consultant
Product governance, engineering prioritization, roadmap discipline. Reports to Gerard Goetz.
AI

Individual TBD

AI Consultant (Incoming)
AI transformation architecture: governance, security, tool standardization, integration strategy.

Leadership

TG

Tony Gonzalez

CEO & Founder
Wants Ramón focused exclusively on Cadmus. Strategic direction and final approvals.
GG

Gerard Goetz

Investor / Executive Advisor
Product governance oversight. Alejandra reports to him.
AR

Adrian Ruiz

Co-Founder / Head of Operations
Operational execution. Active vibe coder (built a 10-hour AI tool replacing 2-week engineering timeline).
JP

Jeff Porter

VP Operations / Product
Building AI projects independently. Active vibe coder.

Open Issues

Organizational and technical

Single Point of Failure

Ramón is the only person who fully understands Cadmus. No documentation beyond Javier's Technical Reference Card.

Engineering Time Under Siege

Ramón gets pulled into non-Cadmus tasks constantly. SOW mandates 70%+ allocation to Cadmus but this requires active protection.

Vibe Coding Sprawl

Alex, Jeff, and Adrian building separate tools using different databases, GitHub repos, and APIs with zero coordination, governance, or security standards.

No Feedback Loop from Clients

Ramón has never talked directly to a brand or client. No follow-up after campaign wrap-ups. No direct market signal on what clients want from Cadmus.

Sales Selling What Doesn't Exist Yet

Sales positions video capabilities to clients, but Cadmus only does metadata-based video classification (not frame-by-frame). Descriptions left intentionally vague.

No Cadmus Health Metrics

No way to measure whether Cadmus is improving or degrading. No A/B testing between versions. No competitive benchmarking.

Consultant & Advisor Landscape

Alejandra (Product Consultant)

Focus: Product governance, roadmap discipline, engineering allocation, tool acceleration.

Mandate: Prioritize Cadmus. Move stalled tools to production. Eliminate manual workflows.

AI Consultant (Individual TBD, Incoming)

Focus: AI transformation architecture. Governance and security for vibe coding. Tool standardization.

Three Phases: (1) Automation (ad ops, RFPs, publisher onboarding), (2) Team adoption, (3) Revenue-generating AI (bid optimization, real-time campaign optimization).

Open question: What is Cadmus V4? Current roadmap items are incremental. Ramón needs protected time and market input to define the next generation.