Startup ecosystems are built, not declared. A decade and 200+ repositories of evidence across six fronts: platforms and registries, access to capital, talent and education, the founder journey, original tools and IP, and trust engineering - all of it inspectable, none of it invented.
A complete investor-discovery SaaS built solo: Express/PostgreSQL/Stripe API with tier-based paywall, React storefront iterated across 286 commits, a six-step waitlist funnel, and a white-labeled support desk - wrapped around a 25,000+ investor dataset.
Sole developer of LuraVPN's identity microservice: FastAPI/PostgreSQL/Redis, JWT/OTP/SSO auth, an influencer referral subsystem, OpenTelemetry + Prometheus observability, a five-layer test suite, and lint/type/test-gated CI shipping Docker images to GHCR.
A polyglot (Go + Python) microservices malware-protection platform: API gateway, threat-intelligence with live Google Web Risk integration, URL filtering, logging/reporting, and notifications - built in ~6 weeks from a self-authored GCP/Kubernetes architecture package.
Tricia Digital Academy: an e-learning platform for digital marketing education - 10 formal design documents (UML, class responsibility, exception policy), then a test-covered NestJS/Prisma/PostgreSQL backend delivered through AI-agent task branches with CI, Husky gates, and Dockerized environments.
Early-stage systems design toward digitizing adoption services for Lagos State's Ministry of Youth & Social Development: a Nigeria-contextual citizen data model (UML), produced in 2021 - my first documented public-sector engagement.
A TypeScript/Express/Mongoose data backbone for a student super-app: loans, jobs, service marketplaces, short video, communities, messaging, and transport booking - ~35 domain models mapping the entire economic life of a Nigerian student.
A full-stack livestock farm-management SaaS maintained over two years: Express/MongoDB API with 13 per-species models, subscription billing, sandbox API keys and audit logging; a Redux-powered React dashboard; and design-first wireframes - run with a small team and CI.
A NestJS/MongoDB financial-system backend - users, accounts, transactions, role-based access - with unit and e2e tests, Sentry/Loggly observability, and a TypeDoc API reference deployed as its own documentation site.
An interactive Python/Jupyter financial model for a Nigerian ISP business: ~15 tunable parameters (NGN pricing, data caps, COGS, capex, location capacity) driving live KPI and break-even visualizations.
A pandas/scikit-learn CLI that turns raw Kuda neobank statements into spending analytics: category and counterparty breakdowns, 3-sigma anomaly detection, and a 30-day cash-flow forecast - built for the messy reality of Nigerian bank exports.
A CI-gated, Terraform-provisioned TypeScript monorepo (Next.js + Express + Supabase) capturing and verifying UK healthcare leads: ~27 test files, append-only RLS-locked migrations, GDPR constraints encoded in the repo, and an AI-agent operating constitution.
A domain-driven TypeScript/Prisma design for the music industry: release campaigns modeled as scoreable real-world 'moments,' with explainable weighted scoring engines, twelve domain entities, and nine numbered design documents - system design as a discipline.
Two 2024 prototypes probing the edges of automation: a Puppeteer-driven tool for discovering town-level communities that match business criteria, and a local GPU voice pipeline (Whisper large-v3 → XTTS v2 voice cloning) for conversational presentations.
Four years (2020–2024) of running a music-mixing brand as a software product: two website generations with a structured SCSS design system, an instrumented booking funnel, designed-then-built APIs, and a lead engine that mined hundreds of real Instagram DMs to decide what to sell.
The most production-grade codebase in this portfolio: a Next.js 16 creative-commerce platform on Cloudflare Workers with 89 test files, 188 commits over 8 months, CI-validated ad-ops configuration, six-platform server-side conversion tracking, and a versioned ad-operations knowledge library.
A Redis-queued Node/FFmpeg system that turns public Reddit content into publishable 16:9 videos - TTS, image compositing, and video assembly as cron-driven backend jobs - shipped with its own hosted JSDoc documentation site.
A media-ops toolbox built to professional packaging standards: rms-scan, a PyPI-published FFmpeg loudness-QC CLI with CI and release automation; an RMS-aware spoken-watermark embedder; Whisper/Vosk shorts pipelines; waveform and PDF stamping utilities.
A Dockerized elizaOS-based AI agent system with hand-authored personas across Discord/Telegram/Twitter - including 'Fred,' a fully-written ideal-customer persona lifted from copywriting study and turned into a conversational agent you can interview.
A tiny Typer/Rich Python CLI that auto-documents the shape of every CSV in a folder with perfectly uniform sibling READMEs - dialect sniffing, deterministic diffable output, proper src-layout packaging and console-script entry points, at 6.5 KB of code.
A consumer-lifestyle API built security-first: XACML-style policy decision/enforcement middleware, granular permission enums, ten validated auth DTOs, privacy-settings modeling, and Loggly-audited access decisions - the trust layer before any feature.
A Lagos mobile-kitchen food-drop launched through a reusable multi-agent AI 'launch factory': Brand, Website, Ads, CRM and Ops agents with reviewer quality gates, generating a full design system and WhatsApp commerce funnel - deployed via Terraform to Cloudflare + Supabase with finance-control guardrails.
An AI meeting-notes pipeline that watches Dropbox folders and writes summaries via OpenAI - proper OAuth refresh flows, ffmpeg chunking around API file limits, multi-tenant config, Docker packaging - while the venture's grander backends were honestly left as named placeholders.
Founder craft, version-controlled: a 31-chapter open summary of Copywriting Secrets published as a free SME resource (with honest AI-authorship disclosure and a human-written branch), plus a personal concept map translating software architecture into startup roles.
A pre-launch media venture treated like a startup: a fully versioned brand thesis, audience architecture, and operating rules in Markdown with a complete logo system - backed by .NET 8 desktop tooling (WPF suite, silent MSI wrapper builder with Authenticode signing) and founder-facing documentation including an investor one-pager.
2019–2024: the posting operation of a solo music brand, fully automated - content curation bots, tweet-screenshot cross-posters, crash-tolerant launchd schedulers with missed-run recovery and idempotency files - the same problem solved four times across two languages as the craft matured.
A browser media-downloader product line: Manifest V3 extensions for Instagram reels and tweet videos, a headless-Chrome backend doing the heavy lifting server-side, and a 2024 extension localized into ~15 languages and wired to a companion product site.
A personal platform play: one Express/Mongoose skeleton harvested from real projects, hardened from 2020 JavaScript into typed, DTO-validated 2024 TypeScript with auth, error handling, and observability baked in - demonstrably reused across at least four shipped backends. Plus a 2020 vendored Huginn, proving the agent-automation obsession predates LLMs.
A two-generation fleet-management API (2020–2021): GPS trip tracking and vehicle records in v1, re-platformed in v2 as a multi-tenant SaaS with companies, plans, API keys, sandbox mode, and guarantors - built by a two-person team, and the origin of a backend boilerplate reused for years.
A conversion-optimized booking site for a Lagos recording studio, shipped in ~9 days: zero-build Tailwind, meticulous Open Graph/SEO metadata, a lead-magnet PDF, and audio voice-note social proof - iterated through AI-agent pull requests.
A JavaFX art-auction platform from late 2017 - sign-up, bidding, history, hand-drawn avatars, 116 commits, full committed Javadoc, and a README that says 'Definitely not production ready.' The university group project that started a 10-year shipping habit.
68 repositories spanning ~2014 to 2026: high-school AP Java and VB.NET, a full UK computer-science degree (Swansea), and labs across four paradigms - imperative, functional, logic, and low-level - preserved end-to-end in version control.
19 repositories of self-built automation (2021–2024): machine bootstrap, dotfiles portability, contact-data pipelines, HTTP test libraries, Keynote remote control - in Python, Ruby, Shell, JXA, and Objective-C.
Seven years of idea-to-repo experiments (2019–2026): startup APIs with real team conventions, a Nigerian land/property dashboard, same-day Valentine apps, LLM probes, and a React task manager with Jest tests. The honest record every founder has and few publish.
Over a gigabyte of curated learning infrastructure in version control: a personal CS library, ideas captured in a self-named .richie file format, other developers' code archived separately from my own, and a living read-later queue - 2021 to 2025.