idosum.github.io
Founder. Builder. Investor.
Africa. AI.
What I'm building
AfricAI Group
AI Advisory & Solutions Integration · Africa
AI advisory and solutions integration for African enterprises — financial services, retail, FMCG, and energy. We diagnose before prescribing: mapping AI priorities specific to a client's operating reality, then identifying and integrating the best global solutions for their context. Co-founded with Oren Salomon.
The engagement platform is built in-house: company intelligence, specialist AI agents per analysis domain, a curated global vendor database, and a knowledge flywheel that captures anonymised learnings from every engagement and compounds them into better outcomes for the next client.
Internal platform and client data are private. The cross-client intelligence layer is AfricAI's core long-term asset.
africai.group ↗Africa VC Intelligence
Proprietary intelligence platform
Structured intelligence on African private companies and investors — built from public data, uniquely collected and structured, with a timeline layer that turns snapshots into a living picture of how companies and capital move across the continent. Covering thousands of companies and hundreds of investors.
Early access available to qualified research partners and investors.
ThinkTok
Edtech · Critical thinking · Ages 9–14
A UK edtech platform teaching critical thinking to students aged 9–14. Built with Galia Shilo Sum — journalist, media literacy educator, and senior editor — who leads content and curriculum.
The content engine and curriculum system are private.
thinktok.ai ↗How I build
All three platforms are built with Claude Code as the primary engineer. Complex features arrive as structured specification files executed by autonomous agents. Each project maintains a CLAUDE.md - a machine-readable brief giving the AI full context on architecture, conventions, current state, and what requires explicit approval before touching.
A cross-project skills library captures operational and intellectual know-how in version-controlled files that load automatically into every session. The knowledge that accumulates across builds is itself part of the system.
All platforms run in production: web and mobile, multiple user types with role-based access, tiered data security with full audit logging of every AI operation, and autonomous background agents. GenAI and agentic AI are combined across the stack - not as separate tools but as one integrated layer. Agents are multi-model: Claude, OpenAI, Gemini, and Perplexity deployed based on what each task requires.
Core stack
Next.js · TypeScript · PostgreSQL · Prisma · Supabase · Turso · Vercel · Railway
AI layer
Anthropic Claude · GenAI + Agentic AI · Autonomous agents · Knowledge loops · Vector search
Build model
Claude Code · CLAUDE.md per project · Skills system · Babysitter autonomous runs · Tiered data security
How I think about Africa
Twenty years of operating in African technology markets - as a founder, investor, and builder across venture capital and public markets - produces a specific kind of knowledge. Not about Africa as a category, but about how individual markets actually work: where capital moves, which regulatory constraints are real versus theoretical, what it takes to build something that lasts in Lagos or Nairobi or Johannesburg versus what the pitch deck version of that looks like.
AfricAI brings together Africa-specific market knowledge, end-to-end strategy and implementation capability, and a curated global vendor network in one place. What makes it compound over time is the cross-client intelligence layer: every engagement generates structured knowledge about which AI approaches work in which African contexts, which vendors perform, which regulatory constraints actually bind. That knowledge feeds the next engagement, and the one after. No firm building generically for global markets accumulates this. It requires being present, committed, and deeply sector-specific across African markets over time.
The Africa VC Intelligence platform applies the same logic to market intelligence: structured, accumulated knowledge about African private companies and investors is genuinely scarce and genuinely valuable. Building it properly takes time and methodology. The timeline layer is what makes it intelligence rather than a snapshot.
Background
20+ years as founder, investor, and builder in African technology markets. Venture capital and public markets across financial services, technology, and media.
Based
London. Focused on Sub-Saharan Africa and select North African markets.
Writing on Africa, AI, and building
Read on Substack ↗