1. Social Trading Platform

   A review directory and social network for active traders

   Live Fintech · Community

   A review directory for brokers, prop firms and funds, fused with a social trading network. It carries a social feed, a copy-trading marketplace that links real MT4/MT5 accounts, a signal-provider marketplace, and an advertising system that lets brokers pay for placement. Three repositories across two build lineages track its evolution from prototype to production.

   • React 18
   • FastAPI
   • MongoDB
   • Tailwind CSS
   • shadcn/ui
   • WebSockets
   • Argon2
   • JWT
   • Google OAuth
   • +7

   Architecture & detail Close

   My role

   • Co-founder, responsible for product, architecture, digital operations and growth
   • Defined the marketplace and monetisation model: CPC/CPM ad placements, sponsored posts, featured listings
   • Authored the paid-growth operating system and attribution architecture
   • Built structured project knowledge so coding agents understand architecture and deployment constraints before touching the app

   What is in it

   • Social feed with direct messages and presence over WebSockets
   • Copy-trading marketplace linking live MT4/MT5 accounts via MetaAPI and CopyFactory
   • Broker, prop-firm and fund reviews with a multi-entity comparison bar
   • Yearly awards system with generated OG images
   • Advertising system: placement wizard, CPM billing scheduler, ad serving
   • Forex calculators, market heatmaps and an economic calendar
   • Admin moderation console for brokers, users, reviews and content
   • Eight languages including Arabic right-to-left

   Architecture

   1. Client

      • React 18 SPA
      • SSR layer
      • i18n · 8 locales · RTL

      Server-rendered for crawlability: the directory is the SEO surface.

   2. Application

      • FastAPI routers
      • WebSocket presence + DMs
      • Ads engine
      • Billing scheduler

   3. Integration

      • MetaAPI / CopyFactory
      • Card + alternative payments
      • Transactional email
      • Object storage

      Payments unified behind one service abstraction rather than three call sites.

   4. Data

      • MongoDB
      • Object storage

   5. Runtime

      • Process supervision
      • uvicorn
      • single-container deploy

      Self-hosted supervision rather than a managed platform.

   Source Confirmed by direct inspection of the running codebase, 2026

2. Organisational Brain

   A knowledge layer that AI agents can actually work against

   Live AI · Knowledge architecture

   A five-layer system that turns fragmented company knowledge into structured context an AI agent can retrieve, reason over and act on. Companies, projects, repositories, technologies, people, compliance material and operating rules are held as linked notes; a graph index makes those relationships machine-readable; a routing layer decides which specialist capability a task needs; and standing rules constrain what an agent may do without approval.

   • Markdown knowledge base
   • Graph index
   • Agent runtime
   • JSON graph
   • Git

   Architecture & detail Close

   My role

   • Designed and built the architecture end to end
   • Authored the routing layer, the search-first procedure and the safety rules
   • Encoded business process, brand standards and deployment constraints as machine-readable rules

   What is in it

   • Knowledge layer: linked company, project, technology and compliance notes with typed frontmatter
   • Graph index over the whole corpus: hundreds of nodes and relationships, clustered into named communities
   • Routing layer mapping an incoming task to the right specialist capability, with explicit anti-patterns
   • A registry of 50 specialist skills, from engineering process to SEO to product discovery
   • Search-first rule: retrieval before questions, so prior work is reused instead of redone
   • Safety rules an agent cannot override: never deploy, never launch spend, never touch production unapproved
   • Resumable state files, so long-running work survives an interrupted session

   Architecture

   1. Knowledge

      • linked entity notes
      • index notes
      • typed frontmatter

      Human-readable. The layer a person actually edits.

   2. Index

      • graph extraction
      • communities
      • incremental update

      Machine-readable, and scoped: a subgraph beats a full-text dump.

   3. Routing

      • task → capability map
      • search-first procedure
      • workflow templates

   4. Capability

      • 50-skill registry
      • engineering · marketing · product · documents

   5. Execution

      • agent runtime
      • local toolchain
      • source control
      • containers

   6. Guardrails

      • no autonomous deploy
      • no autonomous spend
      • review before merge
      • resumable state

   Source The architecture this site’s own graph is modelled on

3. Financial Company A

   Production platform and marketing operations for a regulated CFD brokerage

   Live Fintech · Brokerage

   The public platform for a regulated CFD and forex brokerage, spanning account acquisition, live market data, an introducing-broker programme and a marketing-operations CMS. The build log records more than ninety iterations. It is the group hub for the prop-trading, education and copy-trading sub-brands.

   • React 19
   • CRACO
   • FastAPI
   • MongoDB
   • Microsoft SignalR
   • Tailwind CSS
   • shadcn/ui
   • react-hook-form
   • zod
   • +2

   Architecture & detail Close

   My role

   • Digital, product and technology work across the brokerage operation
   • Product architecture for the signup wizard, account tiers and partner programme
   • Integration of live trading-platform data into customer-facing surfaces
   • Brand system, campaign assets and customer communications

   What is in it

   • Multi-account-type brokerage with a separate social copy-trading tier
   • Live spreads and price streaming over Microsoft SignalR from a liquidity vendor
   • Multi-step OAuth signup wizard validated with react-hook-form and zod
   • Tiered introducing-broker and affiliate partner programme
   • Geo-targeted cross-sell modal, configurable from the admin console
   • Admin CMS covering content, SEO, campaigns, funnel analytics and insights
   • Seven languages including Arabic

   Architecture

   1. Client

      • React 19
      • i18n · 7 locales
      • 35+ routed pages

   2. Application

      • FastAPI routers: accounts · auth · market · campaigns · funnel · webhooks
      • Admin CMS

   3. Market data

      • SignalR price feed
      • live spreads panel
      • economic calendar
      • leverage calculator

      Streamed, not polled: spreads move faster than a request cycle.

   4. Integration

      • Transactional email
      • live chat
      • OAuth providers

   5. Data

      • MongoDB

   Source Confirmed by direct inspection of the running codebase, 2026

4. Property Group

   A property investment business and the deal pipeline that runs it

   Live Property · CRM

   A boutique property business that acquires, redesigns, renovates and resells residential property, with a nineteen-page internal CRM behind the public site, built around the deal. The sourcing side is automated: a search engine discovers listings, a scoring engine ranks them, and the pipeline tracks each one from lead through refurbishment to resale.

   • React 19
   • React Router 7
   • FastAPI
   • MongoDB
   • Search API
   • BeautifulSoup
   • lxml
   • Framer Motion
   • Lenis
   • +2

   Architecture & detail Close

   My role

   • Full build: public site, CRM and the automated sourcing layer
   • Operations modelling: deal pipeline, offer maths, refurbishment costing

   What is in it

   • Property discovery engine with saved searches and a source manager
   • Opportunity scoring engine ranking discovered listings
   • Deal pipeline, offer calculator and portfolio performance with ROI and CSV export
   • Renovation library with a cost estimator supporting custom line items
   • Drag before/after comparison slider driven by a CMS
   • Cinematic parallax hero with smooth scroll and a poster fallback

   Architecture

   1. Discovery

      • search executor
      • source adapters
      • BeautifulSoup / lxml parsing

   2. Scoring

      • opportunity scoring engine
      • saved searches
      • source manager

   3. Pipeline

      • deals
      • leads
      • tasks + notes
      • offer calculator

   4. Delivery

      • refurb library + estimator
      • portfolio ROI / KPIs
      • before/after CMS

   5. Public site

      • marketing pages
      • lead capture → CRM

   Source Confirmed by direct inspection of the running codebase, 2026

5. Copy-Trading SaaS

   Messaging-triggered trade copying onto live accounts, sold as a subscription

   Live Fintech · SaaS

   A subscription service that replicates a signal provider's trades from a messaging channel onto a follower's own whitelisted MetaTrader 5 account. Two-sided: followers subscribe by copy limits and risk controls, providers by the number of channels and followers they can serve. Shipped on iOS and Android alongside the web product.

   • React 19
   • FastAPI
   • Framer Motion
   • Tailwind CSS
   • shadcn/ui
   • RevenueCat
   • MetaAPI
   • CopyFactory
   • i18next

   Architecture & detail Close

   My role

   • Product and subscription architecture across both sides of the marketplace
   • Tier design and packaging for followers and signal providers
   • Admin and operations console specification

   What is in it

   • Live account connection status surfaced through MetaAPI and CopyFactory
   • Subscription and MRR data driven by a mobile billing platform
   • Pricing page rendered from live API data, not hardcoded tiers
   • Operations console: users, providers, subscriptions, connections, mailbox, audit log
   • Transactional onboarding email flow
   • Bilingual marketing site

   Architecture

   1. Signal source

      • messaging channel
      • channel + message metadata only

      Private follower conversations are never read.

   2. Copy engine

      • MetaAPI
      • CopyFactory
      • whitelisted accounts
      • per-tier risk controls

   3. Billing

      • subscription platform
      • follower tiers
      • provider tiers
      • referral rewards

   4. Client

      • React 19 web
      • iOS app
      • Android app
      • two locales

   5. Operations

      • admin console
      • audit log
      • newsletter + mailbox

   Source Confirmed by direct inspection of the running codebase, 2026

6. Prop Firm A

   Funded-account assessments with a live plan builder and a white-label programme

   Live Fintech · Prop trading

   A funded-account business selling simulated assessment programmes that convert into funded trading accounts across forex, crypto and futures, plus a white-label programme letting other brands resell the same infrastructure. The product surface is a configurator: plan type, account size and add-ons priced live.

   • React 19
   • React Router 7
   • FastAPI
   • MongoDB
   • Framer Motion
   • Tailwind CSS
   • shadcn/ui
   • JWT
   • bcrypt
   • +3

   Architecture & detail Close

   My role

   • Product positioning and customer journey
   • Registration, checkout and dashboard purchasing architecture
   • Knowledge-base and customer FAQ architecture
   • Compliance-aware marketing language across the site

   What is in it

   • Live-pricing challenge and plan builder with add-ons
   • Platform comparison across three third-party execution venues
   • Partner and affiliate portal behind an access-code gate
   • Admin CMS with SEO editor, visit and country analytics, and UGC moderation
   • Email-signature generator with a public render tool
   • Installable as a PWA; seven locales with right-to-left support

   Architecture

   1. Client

      • React 19 + Router 7
      • PWA shell
      • i18n · 7 locales · RTL

   2. Configurator

      • plan builder
      • live pricing
      • add-on matrix
      • profit calculator

   3. Admin

      • content + SEO CMS
      • visit/country analytics
      • UGC moderation
      • signature generator

   4. Platform layer

      • third-party execution venues
      • charting embed

   5. Data

      • MongoDB
      • FastAPI service

   Source Confirmed by direct inspection of the running codebase, 2026

7. Trading Software Venture

   Subscription-licensed trading machines, specified rather than mystified

   In build Fintech · Product

   A subscription software business selling a range of expert advisors for MetaTrader 4 and 5, licensed monthly with server-side licence checks, a customer-held kill switch and telemetry from every running instance. The category is built on secrecy and faked backtests, so the product is built on the opposite: a published specification per model, including the worst month and the longest losing streak.

   • MQL5
   • MQL4
   • FastAPI
   • MongoDB
   • React

   Architecture & detail Close

   My role

   • Business model, product range and pricing architecture
   • Brand strategy and identity system
   • Technical architecture: licensing, telemetry and risk enforcement
   • A testing standard defining what an honest backtest has to survive

   What is in it

   • Four deliberately uncorrelated models (trend, mean reversion, breakout and carry) with correlation measured and published
   • An identical spec sheet per model: instruments, cadence, drawdown envelope, worst month, longest losing streak
   • Licence and telemetry server called over a single whitelisted request, with no DLL imports, which prop terminals block anyway
   • Risk presets with a ceiling that cannot be raised, advertised as a feature
   • Prop mode: select a funded-account provider's ruleset and the machine will not breach it
   • A daily morning brief showing what each model did overnight, losses in the same place and size as wins

   Architecture

   1. Terminal

      • MQL5 / MQL4 expert advisors
      • one whitelisted outbound request

      No DLL imports: the classic malware vector, and blocked in managed terminals.

   2. Licence + telemetry

      • account binding
      • subscription state
      • kill switch
      • grace period
      • instance telemetry

   3. Risk enforcement

      • risk presets
      • hard ceiling
      • prop ruleset mode
      • broker-condition guard

   4. Customer

      • model pages with live spec sheets
      • accounts + settings
      • billing
      • downloads

   5. Proving bench

      • live accounts running every model continuously
      • record published permanently

   Source Direction settled 2026; backend slices and terminal core written

8. Payments Provider · Intelligence Portal

   Internal business intelligence for a cross-border payments operation

   Internal Payments · Business intelligence

   A role-gated internal dashboard tracking referral and affiliate commissions, deposits and onboarding status per customer for a multi-currency account and crypto on/off-ramp business. Its interesting half is ingestion: operations receive customer data in whatever format a partner sends it, so the importer takes CSV, TSV, XLSX, JSON and PDF tables and maps the columns itself.

   • React 19
   • FastAPI
   • MongoDB
   • pandas
   • openpyxl
   • pdfplumber
   • reportlab
   • Recharts
   • Tailwind CSS

   Architecture & detail Close

   My role

   • Product definition for the internal operations and sales workflow
   • Ingestion and reporting architecture

   What is in it

   • Multi-format import (CSV, TSV, XLSX, JSON and PDF tables) with an auto column-mapping interface
   • Role gating across admin, staff, sales and affiliate views
   • Referral ranking with per-referrer branded PDF earnings statements
   • Transactions ledger with a profit and fee engine
   • JWT cookie authentication

   Architecture

   1. Ingest

      • CSV · TSV · XLSX · JSON · PDF tables
      • auto column mapping
      • pandas · openpyxl · pdfplumber

   2. Engine

      • profit + fee calculation
      • referral attribution
      • commission ledger

   3. Output

      • Recharts dashboards
      • generated PDF statements
      • CSV export

   4. Access

      • JWT cookie auth
      • admin · staff · sales · affiliate roles

   Source Confirmed by direct inspection of the running codebase, 2026

9. Insurance Brokerage

   Multi-line insurance and wealth brokerage, rebuilt on Next.js

   Live Insurance · Wealth

   An independent brokerage founded in the 1980s, covering seven insurance lines plus a separate wealth and pension vertical. Migrated off WordPress onto Next.js. Every enquiry route (quote, callback, contact, complaint, newsletter) runs through one hardened API endpoint instead of five separate ones.

   • Next.js 15
   • React 18
   • MongoDB
   • Directory mail API
   • Nodemailer
   • Framer Motion
   • Tailwind CSS
   • Radix UI

   Architecture & detail Close

   My role

   • Site architecture and migration from the legacy WordPress build
   • Lead-capture consolidation and content model

   What is in it

   • One generic lead form powering five distinct enquiry types through a single API route
   • Input sanitisation, injection guarding, rate limiting and admin bearer auth on that route
   • Lead notification email over a directory-service mail API with an SMTP fallback
   • Investment and retirement calculators running live compounding maths
   • Multi-column mega-menu with a command-palette site search
   • Installable as a PWA

   Architecture

   1. Client

      • Next.js 15 App Router
      • mega-menu nav
      • site search modal
      • PWA

   2. Content

      • site config
      • services config
      • wealth config

      Brand and service content held as config, not scattered through components.

   3. Lead API

      • single route
      • sanitisation
      • injection guard
      • rate limit
      • bearer auth

   4. Notify

      • directory-service mail API
      • SMTP fallback

   5. Data

      • MongoDB native driver

   Source Confirmed by direct inspection of the running codebase, 2026

10. Growth Operating System

    A paid-growth system with the attribution designed before the spend

    Live Growth · Analytics

    A documented growth operating system running the loop create → launch → measure → analyse → learn → improve, optimised against the real funnel from impression through to retained user, not vanity metrics. It began with a read-only tracking audit, which found the measurement broken in four specific ways before a single unit of spend was committed.

    • GA4
    • Tag manager
    • Conversions APIs
    • Measurement Protocol

    Architecture & detail Close

    My role

    • Designed the growth operating system and its knowledge structure
    • Ran the tracking audit and specified the attribution architecture
    • Wrote the safety rule preventing autonomous campaign or budget changes

    What is in it

    • Tracking audit found duplicate-firing analytics, pixels with no IDs set, and a consent banner gating nothing
    • Attribution chain specified end to end: platform → campaign → creative → click → landing → registration → activation → retention
    • First-party capture of UTMs and ad-platform click IDs, persisted and carried into the registration payload
    • Server-side funnel events keyed on a shared event ID so each platform dedupes against its own pixel
    • Delayed conversion upload for clicks that convert outside the session
    • Event, UTM and KPI taxonomies defined before campaigns exist

    Architecture

    1. Capture

       • UTM parameters
       • ad-platform click IDs
       • first-party storage · ~90 days

    2. Carry

       • first-touch + last-touch snapshot
       • declared fields on the user model

       Declared fields, not a raw request-body fallback.

    3. Dispatch

       • server-side events
       • platform conversions APIs
       • shared event id

    4. Reconcile

       • offline conversion import
       • consent gating
       • dedupe against client pixels

    Source Tracking audit and attribution architecture, 2026

11. Brand & Campaign Systems

    Identity, campaign and collateral systems across a group of brands

    Live Brand · Design

    Visual identity and marketing systems across brokerage, prop trading, payments, broker technology, insurance, education and review brands. Full identity kits with light, dark and gradient variants; campaign asset sets; award badge systems; exhibition builds; and a design-token file per brand, so a look can be ported between projects deliberately instead of by copying CSS.

    • Illustrator
    • SVG
    • Design tokens
    • Vector production workflow

    Architecture & detail Close

    My role

    • Brand development and visual identity
    • Campaign and collateral design
    • Design-token systems feeding the front-end builds

    What is in it

    • Identity kits across eight brands spanning brokerage, prop trading, payments, education and review platforms
    • A brand guideline system defining product sub-domains, each with its own gradient variant
    • A market-entry campaign for a broker-technology brand: isometric illustration set, twelve-icon system, light and dark logo variants
    • An award badge programme used to endorse ranked brokers
    • A full trade-show exhibition build: back wall, counter, fascia panels and stationery
    • Per-brand design-token files carried into the codebases

    Source Asset libraries and per-brand token files across the group

12. Share-Sale Pitch Deck

    A corporate pitch deck as an encrypted single-page site

    Live Payments · Corporate

    A password-protected landing-page pitch deck built to support the sale of a majority stake in a payments entity. Rather than gating the page in JavaScript, which protects nothing because the content has already been shipped to the browser, the content is encrypted client-side with AES-GCM and decrypted only once the correct passphrase is entered.

    • HTML
    • CSS
    • JavaScript
    • Web Crypto API

    Architecture & detail Close

    My role

    • Built to brief for the commissioning stakeholder
    • Content, structure and security model

    What is in it

    • Client-side AES-GCM content encryption, not a cosmetic password gate
    • Static HTML, CSS and JavaScript, so there is no server to compromise
    • Product and app mockups rendered as animated CSS

    Source Delivered 2026

The tool underneath matters far less than the architecture on top of it. What transfers to another company is the structure, the retrieval design and the rules, not the note-taking app. See the full system →

Hover a count to see which systems it refers to.

1. 01

   Take the whole problem

   The compliance wording, the funnel and the data model are usually one decision seen from three chairs. Splitting them across three teams is how products end up internally inconsistent.

2. 02

   Retrieve before asking

   Most questions a business asks have already been answered somewhere inside it. Search first, ask second, or the same research gets paid for twice.

3. 03

   Claim only what is built

   A specification you can check beats a promise you cannot. Where something is unproven, it gets labelled unproven, including on this site.

4. 04

   Instrument before spending

   Attribution designed after launch produces numbers nobody trusts. The measurement architecture is part of the build, not a phase after it.

5. 05

   Automate with a brake

   Anything irreversible gets an approval gate: a deploy, a spend, a delete. Autonomy is worth having right up to the point where a mistake becomes permanent.

6. 06

   Run it, then say it works

   Generated source is not a working system. The claim comes after the thing has actually been run, and if it failed, that gets said plainly.

1. 2025–present

   AI-native systems & organisational intelligence

   Independent · applied across the group

   Designing the layer that lets AI systems work against a company's real knowledge instead of a blank context window.

   • Built a five-layer organisational brain: linked knowledge, a machine-readable graph index, a routing layer, a specialist capability registry and an execution runtime
   • Encoded business process, brand standards, deployment constraints and safety rules as machine-readable instructions
   • Established the operating discipline: retrieval before questions, verification before claims, approval gates on anything irreversible
   • Applied it to live product work: project-aware coding agents, growth analysis, brand development and documentation

   • AI
   • Knowledge architecture
   • Context engineering
   • Automation

2. 2024–present

   Founder & product lead

   Co-founded ventures · fintech products

   Taking trading and fintech products from concept to production: architecture, monetisation, growth and brand together.

   • Co-founded a broker and prop-firm review directory fused with a social trading network and a copy-trading marketplace
   • Designed the advertising and monetisation model, then the attribution architecture that would make the spend measurable
   • Scoped a subscription expert-advisor business built on published specifications, server-side licensing and enforced risk ceilings
   • Ran brand strategy and identity for both, including the category anti-patterns each brand had to avoid

   • Product
   • Fintech
   • Growth
   • Brand

3. 2019–present

   Digital, product & technology across an FX and fintech group

   An FX, prop-trading and payments group

   Broad responsibility across the digital, technology, product and marketing operation of a brokerage group and its sub-brands, spanning five CFD and FX brokerages, three asset management firms, a prop-trading platform and a payments business.

   • Production brokerage platform: account tiers, signup wizard, live spreads over a streaming price feed, partner programme, marketing-operations CMS
   • Prop-trading product: live-pricing plan builder, platform comparison, partner portal, compliance-aware marketing language
   • Copy-trading subscription product across web, iOS and Android, with an operations console behind it
   • Internal payments BI: multi-format ingestion, commission ledger, branded PDF statements
   • Integration of live MetaTrader and MT Manager data into customer-facing surfaces through APIs and sockets

   • Fintech
   • Product
   • Engineering
   • Operations

4. 2019–present

   Work outside financial services

   Property, insurance and wealth clients

   The same method applied to unfamiliar industries, which is the real test of whether a method is a method.

   • Property investment business: public site plus a nineteen-page deal-pipeline CRM with automated sourcing, opportunity scoring and refurbishment costing
   • Insurance and wealth brokerage: migration from WordPress to Next.js, five enquiry types consolidated onto one hardened API route, live financial calculators
   • Each rebuilt from the operating model outward, not from a template inward

   • Product
   • Engineering
   • Operations

5. 2009–present

   Brand, design & digital strategy

   Across the group and client work

   The continuous strand underneath everything else: identity, campaign systems and the digital strategy around them.

   • Identity systems across brokerage, prop trading, payments, broker technology, education and review brands
   • Campaign systems including a market-entry campaign with an isometric illustration set and a twelve-icon system
   • Award badge programmes, exhibition builds and full collateral systems
   • Digital strategy, SEO, CRM, email infrastructure, conversion and funnel work alongside the design

   • Brand
   • Design
   • Growth
   • SEO

6. 2006–2009

   MA Graphic Design

   University of the Arts London

   Graphic design, visual communication, branding, typography, user experience and strategic design.

   • The foundation the rest was built on, later extended into web technology, product, fintech and AI

   • Design
   • Typography
   • Brand