Home

The pipeline

How Ffembi works

Ffembi doesn't just list names — it traces relationships. Every node and edge in the graph is discovered, extracted, verified, and continuously revalidated.

From discovery to graph

1. Discovery

The Ffembi team identifies entities across the ecosystem — startups, funders, ESOs, events — through news mentions, blog articles, media coverage, and social signals.

2. Extraction

AI-powered scrapers pull structured data where possible. For hard-to-reach surfaces like LinkedIn, human researchers extract manually — every fact tied to a source.

3. Verification

Human reviewers cross-check AI extractions against original sources. Each relationship and attribute is validated before it enters the graph — no unverified claims.

4. Graph

Validated entities and relationships are written to SurrealDB as typed nodes and edges, guided by our evolving schema. The map comes alive.

Machines scale, humans verify

AI-powered scraping

  • Automated extraction from public websites, news outlets, and directories
  • Entity recognition and relationship inference from unstructured text
  • Continuous crawling to detect new entities and updated attributes

Human verification

  • Manual extraction from gated or complex sources, e.g. LinkedIn
  • Cross-referencing every claim against its original source
  • Source URLs recorded for every relationship in the graph

All data is normalized against our evolving schema before entering SurrealDB.

Data integrity

Source-defined edges

Relationships are not crowdsourced guesses. Every edge is backed by a referenced source — a news article, funding announcement, or verified public record.

Monthly revalidation

Ecosystem data decays fast. Sources are revisited every month. Stale relationships are flagged, updated, or retired.

SurrealDB backend

Nodes and edges are first-class citizens, not relational workarounds — fast traversals, flexible schema, live queries.

Node types

Every node in the graph is one of six types, each with its own subtypes. This is what every entity gets classified against before it's written to the graph.

TypeSubtypeDescription
PersonAny individual human tracked in the graph, regardless of affiliation — the agency layer; organizations act through people.
EntityBuilderEntrepreneurial venture or startup actively building a product/business, idea to scale.
FunderOrganization providing capital — grants, investment, donor funding.
EnablerESO delivering programmatic support — accelerators, incubators, hubs, mentorship.
KnowledgeUniversity, research body, or academic institution contributing research, talent, or knowledge.
ConnectorNetwork, community, or association focused on linking people and organizations.
RegulatorBureau/agency responsible for compliance, registration, or oversight.
GovernmentMinistry or state body engaged in policy-making and economic planning.
ResourceFinancialMoney in any form — grants, equity, debt, revenue.
KnowledgeIntellectual assets — research, curricula, frameworks, market data.
DigitalSoftware, platforms, data, or digital infrastructure.
PhysicalTangible assets — equipment, office space, raw materials.
Human CapitalLabor, skills, or expertise — mentors, seconded staff.
EventCompetitionJudged contest, e.g. pitch competition.
GatheringInformal or social meetup.
LearningWorkshop, training, or educational session.
ConferenceLarger multi-session themed gathering.
ShowcaseDemo day or exhibition.
ProgramSupportNon-financial structured support — accelerator cohort, mentorship.
FundingStructured capital deployment — grant cycle, investment round.
PolicyGovernment/regulatory initiative to shape ecosystem conditions.
LocationPhysicalSpecific physical address or place.
OnlinePurely digital environment.
HybridCombination of physical and online.

Sectors

Every entity is tagged against a fixed sector list defined. We add, rename, and retire sectors as our research on the ecosystem evolves — this isn't a static taxonomy.

agritechfoodtechmanufacturing_industrialtechlogistics_supply_chain_techmobility_transporttechfintechai_data_big_techcybersecurity_data_privacyecommerce_retailtechhealthtechedtechhr_tech_future_of_workenergy_cleantechwatertech_sanitationclimatetech_green_economyproptech_real_estate_techgovtech_civictechlegaltechmedia_creator_economy_cultural_techtraveltech_tourismtech_hospitalitybiotech_life_scienceswaste_management_circular_economy_techcredit_facilities_asset_financing

Edge dictionary

Every relationship type is declared once with a constraint on which node types can connect. The schema itself, not convention, is what stops a nonsensical edge from being written.

TRANSACTION

awardedgrantedinvested infunded

ENGAGEMENT

attendedpitched atalumni offeatured inparticipated infacilitatedran

GENERATION

implementedfoundedspun off fromcreated

CONNECTION

mentoredpartnered withpart ofhostssupportedmember ofboard member of

LOCATION

located at

Claim your node — coming soon

Every entity in the graph gets a public profile — contact info, pitch decks, social links. Profiles are seeded by our research team today; the goal is for you to claim and maintain your own.

Editable basics

Claim your profile to edit bio, contact details, links, and descriptions.

Source-locked relationships

Edges — who funded you, who you mentor, where you're based — stay defined by verified sources to preserve graph integrity.

Aspirations

We are constantly iterating - however, existing ecosystem platforms inspire and guide the vision for Ffembi. We hope to build some kind oc chimera based on them.

VC4A & F6S

Existing platforms for organizing startup ecosystems online — profiles, programs, opportunities. We build on their instincts and add the graph layer.

Wikipedia

Reference point for governance: open enough to be community-maintained, structured enough to stay trustworthy.

Linktree

Reference point for the claimed profile — one simple, self-maintained link-in-bio page per node, instead of a directory entry someone else has to keep updating.

PitchBook

Reference point for depth: structured, verifiable data on deals, funding, and relationships that people actually rely on to make decisions — the level of trust we want the graph to eventually earn.

Built with

SvelteKit · SurrealDB

References

The schema isn't invented from scratch — it's assembled from existing entrepreneurship ecosystem research and products, and revised as we read more.

Isenberg's entrepreneurship ecosystem domains

Foundational domain model (policy, finance, culture, support, human capital, markets) behind how we categorize actors.

Understanding entrepreneurial ecosystems through social network analysis (SNA) - 2012

A research activity that used social network analysis to map entrepreneurial ecosystems in Uganda — the closest existing precedent for Ffembi.

Kauffman Foundation's Entrepreneurial Ecosystem Framework

Focuses on cultivating inclusive, connected, and resource-rich environments where startups can thrive - removing systemic barriers, aligning community assets, and empowering ecosystem builders—rather than relying on top-down economic mandates.

ANDE Entrepreneurial Ecosystem Diagnostic Toolkit

Aspen Network of Development Entrepreneurs' field methodology for assessing an ecosystem — informs our stakeholder categories.

GIZ — Guide for Mapping the Entrepreneurial Ecosystem (2018)

GIZ’s Observe–Analyse–Visualise methodology for entrepreneurial ecosystem mapping, itself built on Isenberg and ANDE.

Mason & Brown's actor-based ecosystem framework

Treats people, not just organizations, as the primary unit of an ecosystem — why Person sits at the center of our schema.

Last updated: 27 July 2026

Schema version evolves continuously. Sources revalidated monthly.