Powered by Fermia

One Platform. One Data Layer. Nine Applications. Real-Time Everything.

“The technology disappears. The value stays.”

1
Platform
1
Data Layer
9
Applications
<100ms
Latency

Section 01 Architecture Overview

Five layers, one platform. Every ECA application reads and writes from the same data layer. No ETL, no sync jobs, no duplicate data.

User Interfaces
Atlas Portal Compliance Matchmaking Events Academy Governance Public Site
Application Layer
Membership CRM Accounting BPMN Chat
Intelligence Layer
AI Agents (Spark) MCP Tools Semantic Search
Data Layer
OODB (Dynamic Schemas) Taxonomies Knowledge Graph
Infrastructure
SpacetimeDB WebSocket Real-Time Sync

Section 02 One Object, Nine Views

The digital twin thesis: a single Farm object in the OODB is the source of truth. Every application renders its own view of the same data — no duplication, no sync.

Farm Object
name: Kochere Station
gps: 6.16°N 38.2°E
altitude: 1,850m
area: 2.3 ha
variety: Heirloom
process: Washed
Atlas
Pin on map with altitude, variety, and terroir data
Compliance
First node in EUDR traceability chain
Member Profile
“My Farms” list item with area and production data
Matchmaking
Sourcing match filter by variety, altitude, process
Events
Lot origin attribution for cupping events
Tourism
Farm visit destination with GPS and gallery

→ When an exporter updates their farm data, every application sees the change instantly.

Section 03 OODB Taxonomy Tree

Eight purpose-built taxonomies encode the structure of Ethiopia’s coffee sector — regions, varieties, processing methods, certifications, and more.

Fermia OODB — Ethiopian Coffee Sector
CoffeeRegion (12 types)
Sidamo → Yirgacheffe, Guji → Hambela, Jimma, Harar, Limu, Kaffa, Wellega, Bench Maji, Illubabor, Gedeo
CoffeeVariety (10 types)
Heirloom → Kurume / Dega / Wolisho / Gesha | JARC → 74110 / 74112 / 74158 / 74165
ProcessingMethod (4 types)
Natural, Washed, Honey, Anaerobic
CertificationType (6 types)
EUDR, Organic, Fair Trade, Rainforest Alliance, SCA, CQI
MemberCategory (5 types)
Exporter, Farmer, Cooperative, Washing Station, Roaster
CourseCategory (5 types)
Cupping, Export Procedures, EUDR Compliance, Agronomy, Barista + Q-Grader sub
EventType (5 types)
Addis Coffee Cup, Coffee Diplomacy, Stakeholder Forum, Trade Fair, Coffee Tourism
ComplianceRequirement (5 types)
Geolocation, Deforestation Check, Due Diligence Statement, Traceability, Legal Compliance

Section 04 Real-Time Capability

SpacetimeDB is not a database with an API bolted on — it is a stateful compute engine that pushes every change to every connected client over WebSocket.

WebSocket Native

All clients receive updates instantly via persistent WebSocket connections. No polling, no webhooks, no delayed sync.

Live Auction Bidding

Bidding events with <100ms latency. Every participant sees every bid as it happens — fair, transparent, real-time.

Live Dashboards

OPI scores, member counts, compliance percentages — all update as data changes. No refresh button needed.

Instant Availability

“When an exporter updates their lot availability, every buyer browsing the Atlas sees it immediately.”

Section 05 AI Agents

Spark agents use MCP tools to read and write OODB data. The platform IS their working memory — they don’t need external databases, APIs, or data stores.

Matchmaking Agent
Processes sourcing briefs from international buyers. Queries OODB for farms matching variety, altitude, processing method, and volume requirements. Ranks matches by fit score and generates introduction emails.
Compliance Agent
Validates EUDR documentation packages. Flags missing geolocation data, incomplete due diligence statements, and expired certifications. Auto-generates compliance reports for submission.
Sector Report Agent
Assembles quarterly sector reports from live platform data — export volumes, price trends, member growth, compliance rates. No manual data collection needed.
Member Service Agent
Answers member questions in Amharic and English. Guides new members through onboarding, explains certification requirements, and helps navigate the platform.

Key insight: Each agent uses MCP tools to read and write the OODB — the platform is persistent working memory for AI, not just a database for humans.

Section 06 Why Not Alternatives?

The conventional approach — Salesforce for CRM, Shopify for commerce, custom code for compliance — creates data silos, integration tax, and vendor lock-in.

Capability Fermia Salesforce + Shopify + Custom
Single data layer One OODB 3+ databases + ETL pipelines
Real-time WebSocket native Polling / webhooks / batch sync
Dynamic schemas OODB — no code required Custom development per schema change
AI agents MCP integrated, native access Separate AI layer, API bridges
Cost (5 years) ~$50K/year ~$200K+/year
Vendor lock-in Open source stack Proprietary platforms
Offline-capable Client-side data cache Cloud-dependent
Ethiopia data sovereignty Self-hosted option Data in foreign jurisdictions

Section 07 Phased Rollout

Technology deployment aligned to the B³ project portfolio — each phase delivers operational modules that unlock the next.

Phase 1
2026
  • Membership Portal
  • Coffee Atlas
  • Compliance Portal
  • OODB + Taxonomies
Phase 2
2026–2027
  • B2B Matchmaking
  • Coffee Academy
  • Events Platform
  • CRM + Accounting
Phase 3
2027
  • Full Platform Integration
  • AI Agents (Spark)
  • Semantic Search
  • Live Auction
Phase 4
2028+
  • Coffee Tourism
  • Industrial Production
  • Advanced Analytics
  • API Ecosystem