⬡ Platform Architecture

The Foundation Intelligence Layer.

The biological intelligence infrastructure that sits between raw biological data and decision-making — making equine biology computable.

Explore Capabilities API Documentation →
L4 Applications JumpHealth™ · JumpID™ · JumpChain™
L3 Core Engines Foundation Intelligence · HorseVector · Knowledge Graph
L2 Platform Website · Dashboard · API · SDK
L1 Corporate Identity Mission · Values · Story
Foundation Intelligence

Five core capabilities of
biological intelligence.

Biological Intelligence is the structured representation of biological systems that enables computation, reasoning, and evidence-grounded decision-making. Foundation Intelligence is its implementation for equine biology.

Representation

Unified encoding of biological entities. Every gene, variant, phenotype, and breed becomes a computable object.

→ HorseVector embeddings

Integration

Multi-source data fusion. Genomes, phenotypes, literature, and field data unified into coherent structure.

→ Knowledge Graph pipeline

Reasoning

Biological inference over structured knowledge. Inferring relationships, classifying variants, and predicting outcomes.

→ Equine Foundation Model

Traceability

Every output links back to evidence. Full provenance chain: source paper, sample ID, genomic coordinate.

→ Radical Traceability system

Prediction

Decision-ready outputs. Risk scores, mating recommendations, variant classifications — with confidence intervals.

→ Decision Intelligence layer
HorseVector

The equine biological embedding layer.

HorseVector encodes biological entities — genes, variants, phenotypes, breeds — as dense, computable vectors. Just as word embeddings revolutionized NLP, HorseVector aims to transform how equine biology is represented and computed.

  • Each entity is encoded as a high-dimensional vector capturing its biological relationships and properties
  • Similar biological entities cluster in vector space, enabling analogical reasoning across genes and phenotypes
  • Pre-trained on curated equine genomic data from 17 breeds and 500K+ variants
  • Designed to be adopted as an open standard for equine biological representation

Our goal is to establish HorseVector as the default coordination layer for equine biological representation.

HorseVector Embedding Space ● Active
Gene MSTN (Myostatin) [0.12, -0.87, 0.43, ...]
Variant g.229786C>T [0.31, -0.64, 0.18, ...]
Phenotype Speed Performance [0.28, -0.71, 0.35, ...]
Breed Thoroughbred [0.19, -0.52, 0.61, ...]
Knowledge Graph

Gene → Variant → Phenotype → Breed.

The EQAI Knowledge Graph connects biological entities through validated relationships. Every edge represents curated, cross-referenced evidence — not statistical correlation alone.

→ has_variant → associated_with → manifests_as → prevalent_in → validated_by → contradicts
79
Diseases
14
Pathogenic Variants
17
Breeds Mapped
🧬
MSTN
🧬
DMRT3
🧬
GSDMC
⚡
g.229786C>T
⚡
rs114018383
◈
Speed
◈
Gait
◆
Thoroughbred
◆
Icelandic
Gene
Variant
Phenotype
Breed
Foundation Model

The equine-specific foundation model.

We are building the first comprehensive foundation model pre-trained on curated equine biological data. It aims to reason over genes, variants, phenotypes, and breeds — producing hypotheses for expert validation, never conclusions.

Our goal is to build what others build on — not replace scientific review, but augment it with structured biological reasoning.

🎯

Equine-Specific

Trained on horse biology, not adapted from human models

🔬

Evidence-Grounded

Every inference links back to Knowledge Graph evidence

📊

Confidence-Aware

Outputs include confidence intervals and validation status

🔄

Continuously Updated

New evidence triggers re-assessment and model updates

Training Pipeline ◐ In Progress
01
Data Curation
17 breeds · 79 diseases · 14 variants
02
HorseVector Pre-training
Biological embedding space
03
Knowledge Graph Alignment
Entity-relation embedding
04
LoRA Fine-tuning
Qwen 3.5 equine Q&A
05
Cross-Breed Validation
Independent population testing
06
Expert Review
Veterinary geneticist validation
APIs & Data Layer

Query by breed, gene, variant,
phenotype.

RESTful APIs and GraphQL endpoints providing programmatic access to the entire EQAI knowledge base. Build on the foundation layer.

GET /v1/breeds List all genotyped breeds
GET /v1/genes/{symbol} Gene details + variants
GET /v1/variants/{rsid} Variant with OMIA + evidence
GET /v1/phenotypes/{id} Phenotype associations
GET /v1/breeds/{id}/frequencies Allele frequencies per breed
POST /v1/screen Genetic risk screening
GET /v1/knowledge-graph/query Graph traversal queries
POST /v1/horsevector/embed Entity embedding generation
◈ GraphQL endpoint: /graphql — introspection enabled
REST — cURL Copy
# Query a gene's variants and phenotypes curl -X GET \ https://api.eqaios.com/v1/genes/MSTN \ -H "Authorization: Bearer $TOKEN" # Response { "symbol": "MSTN", "name": "Myostatin", "chromosome": "ECA18", "variants": [3], "phenotypes": [ "speed_performance", "muscle_development" ], "breeds_affected": 17, "validation": "✅ validated" }
GraphQL Copy
# Query the Knowledge Graph query { gene(symbol: "DMRT3") { symbol name variants { rsid effect confidence } associatedPhenotypes { name evidence_level } breedFrequencies { breed allele_frequency } } }
Evidence Model

From observation to decision.
And back again.

The Evidence Model is not a pipeline. It is a cycle. Every decision produces an outcome. Every outcome becomes new observation. The system learns, the model updates, truth is versioned.

01

Raw Observation

Samples · Veterinary records · Field data · Research papers

↓
02

Evidence

Genomes · Phenotypes · Variants · Published research

↓
03

Validation

Cross-referenced · Peer-reviewed · Literature-confirmed

↓
04

Knowledge

Curated, connected, validated — the living Knowledge Graph

↓
05

Foundation Intelligence

Reasoning over structured knowledge — the equine-specific intelligence layer

↓
06

Prediction

Risk score · Mating recommendation · Variant classification

↓
07

Decision

Breeding choice · Treatment plan · Racing strategy · Conservation action

⟳
Feedback Loop: Every decision produces an outcome in the real world. Outcomes become new observations — fed back into the system. The Knowledge Graph enriches, models retrain, and truth is versioned.

This is what makes EQAI living infrastructure — not a static database.

Build on the foundation layer.

Access the EQAI API, explore the Knowledge Graph, or integrate HorseVector embeddings into your research workflow.

Request API Access Read Documentation

Decode to Decide.