FBMacagba

London based digital product builder

Ferdinand Macagba

AI-assisted application developer with product judgment, responsible AI discipline and live shipped work.

I research the problem, define requirements, specify the workflow and direct every revision. AI assists implementation. I own the product decisions, review and release.

Download CVReview shipped workStart a conversation

About Me

From Operations to AI-Assisted Application Development

I spent over a decade in hospitality and remote operations, managing e-commerce fulfilment, coordinating distributed teams, and solving real business problems under pressure. That career taught me what organisations actually need from software: tools that work reliably, workflows that reduce friction, and systems built for the people who use them daily.

Since 2023, I have been building production applications using a structured, human-led process: I research the problem, define requirements, design the workflow, and specify every decision. AI assists implementation; I own the product. This means every application starts with a PRD, progresses through defined stages, and ships only after review, testing, and iteration.

I take responsible AI seriously as engineering discipline. Every application I build applies input validation, human approval gates, output verification, and least-privilege permission controls. I hold a certification in Managing AI Governance with ISO 42001 and a Diploma in Applied Generative AI.

Current focus: SaaS platforms, e-commerce and AI-enabled applications, built for real users and real business problems.

10+

Years Client Experience

6

Live Projects Deployed

10+

Tech Certifications

London

Based in UK

My Process

My Development Process

AI assists research, planning and implementation. I remain responsible for the product decisions, requirements, workflow, testing, risk controls and release approval.

How the work flows

AI assists the build; I own every decision and the release gate. Each loop ends in a human approval before anything ships.

  1. PRD / Spec

  2. AI-Assisted Build

    AI implements

  3. Human Review

    ◆ approval gate

    reject → revise build

  4. approve

  5. Test & Verify

  6. Ship

  7. iterate → next PRD / Spec
Human-led AI-assisted
Input validationOutput verificationLeast privilegeHuman in the loop
01

Problem Research

02

Idea Development

03

Product Planning

04

Requirements (PRD)

05

Prompt Engineering

06

Workflow Definition

07

AI-Assisted Build

08

Review & Testing

09

Iterative Development

10

MVP Approval

11

Feature Planning

12

Application Lifecycle

13

Responsible AI Controls

14

Input Validation

15

Human Approval Gates

16

Version Control & Continuity

Process in Practice

Case Study: BytesCRM

AI-Powered CRM with Embeddable Chatbot

ProblemSmall businesses lose website visitors without capturing them as leads. The tools that exist are either too expensive (Salesforce, HubSpot at scale), too generic (no built-in lead capture), or disconnected (chatbot, CRM, and billing are three separate products stitched together).
ResearchIdentified three gaps: solo founders cannot afford per-seat enterprise pricing; standard CRMs have no website lead capture built in; and the setup required is too complex for a business owner without a developer.
Requirements (PRD)Embeddable chatbot widget (single script tag), lead inbox with overflow-and-unlock model, CRM pipeline (contacts, deals, tasks), autoresponder email sequences, Stripe-powered tiered billing with pay-per-lead unlocks, and role-based access (User, Admin, Developer).
Architecture DecisionMulti-tenant SaaS: each business account has isolated data, its own chatbot configuration, and independent billing. React 18 frontend, Base44 (Deno-based serverless) for the backend and entity store, Stripe for subscriptions and per-lead payments, Resend for transactional email, OpenRouter with Claude and Gemini powering the chatbot AI. The chatbot ships as a standalone embeddable JS widget, separate from the main app bundle.
AI-Assisted ImplementationEach module scaffolded with AI assistance, output reviewed, gaps identified, corrections specified, then retested. Human sign-off at each module before moving on. AI provider cost monitored via a built-in pricing governance dashboard.
Responsible AIChatbot conversations are rule-scoped: AI cannot take actions outside its defined flow. Role-based access enforced throughout (User, Admin, Developer). Stripe environment switching (Test vs Live) requires explicit developer-level action with full audit logging. Overflow lead expiry emails give business owners 48h and 24h warnings before leads are permanently removed.
IterationFive real pain points drove post-launch additions: losing website visitors (added chatbot widget); cost barrier (free tier with overflow model); attribution blind spots (UTM, Google Ads click ID, Meta Pixel tracking); onboarding friction (guided setup, under 10 minutes to first lead); and account recovery (structured deletion cooldown and re-registration flow).
MVPChatbot widget + lead inbox + CRM pipeline + Stripe billing deployed. Tiered plans (Free, Starter, Pro, Business, Enterprise) live. Smart Chatbot AI add-on and message top-up packs in the roadmap.

Case Study: Active Paw

Demand-Driven Shopify Store with AI Backend

ProblemA pet accessories store requiring constant manual effort: sourcing from suppliers, writing listings, managing prices, and retiring dead stock. The owner wanted customer demand to decide what gets listed, not manual buying decisions.
ResearchIdentified that Shopify's native search gap (fewer than 12 results) was wasted demand signal. Real visitor searches were revealing products the store didn't stock but suppliers could fulfil, and the opportunity was to capture that signal and act on it automatically.
Architecture DecisionTwo-repo system: Shopify Liquid theme as the demand surface, PawScout Cloudflare Worker as the intelligence layer. All supplier calls, quality gates, pricing, and import logic live in the worker; the theme is stateless. Shopify App Proxy used throughout so PawScout outages degrade gracefully to standard search.
AI IntegrationLLM quality gates on every ghost card: pet-relevance check prevents off-category products appearing in backfill. Vision API vets product images before caching, ensuring ghost cards show the correct hero image rather than a mismatched supplier thumbnail.
Responsible AIPet-relevance guard added after an LLM outage let kitchen tools slip through quality gates, so the store now stays stocked with genuine pet products regardless of LLM availability. Ghost cards only surface quality-gated results; if the worker is down, the store runs as plain Shopify search with no errors shown.
IterationThree real bugs fixed post-launch: ghost card showing a cat image while the product showed a dog (fixed by caching vision-vetted hero image in PawScout); modal loading slowly on first open (fixed by pre-caching supplier data on search, not on click); off-category products slipping through during LLM outages (fixed by worker-level pet-relevance fallback).
MVPSearch backfill with ghost cards, import-on-first-interest (product live within seconds of first click), 30-day performance cull, and fulfilment automation. Live at activepaw.co.uk.

Portfolio

Featured Projects

Human-led, AI-assisted applications built through structured requirements, workflow design and iterative development.

Case Studies

BytesCRM

AI-Powered CRM with Embeddable Chatbot Widget

Problem

Small businesses lose website visitors without capturing them as leads, then manage the few they do get across disconnected tools: a separate chatbot, a separate CRM, and separate billing.

Solution

A single platform: an embeddable AI chatbot widget that captures leads from any website, a full CRM pipeline to manage them, and a pay-per-lead overflow model so businesses start free and only pay as they grow.

Technical highlights

Multi-tenant SaaS architectureReact 18 + Base44 serverless (Deno)Embeddable JS chatbot widgetStripe subscriptions + pay-per-lead billingOpenRouter / Claude / Gemini AI

The overflow lead model was a deliberate product decision: leads above the plan limit are locked behind a one-time unlock payment rather than lost or auto-billed. This kept the free tier genuinely useful while creating a natural, low-friction upsell path.

Active Paw

Demand-Driven Pet Store: Shopify Theme + AI Backend

Problem

A pet store where every product listing required manual effort: sourcing from suppliers, writing descriptions, managing prices, and retiring dead stock. The owner wanted customer demand to drive the catalogue automatically.

Solution

A two-repo system: a Shopify Liquid theme that captures visitor demand via search backfill and ghost cards, paired with PawScout, a Cloudflare Worker that runs LLM quality gates, vision-vets images, and auto-imports products the moment a real visitor shows interest.

Technical highlights

Shopify Liquid theme + App ProxyCloudflare Worker (PawScout)LLM quality gates + vision-vetted imagesImport-on-first-interest pipeline

Built on Shopify App Proxy so PawScout can fail completely without breaking the store; all calls degrade to no-ops. The theme is deliberately stateless on quality; all intelligence lives in the worker.

All Projects

BytesAI Learn

Curated Discovery Platform for Free AI Courses

A human-curated platform surfacing 100+ verified-free AI courses from Google, Anthropic, Stanford, DeepLearning.AI, Microsoft, NVIDIA, and 20+ other providers. Curation is the product: every course verified free, precisely categorised, and contextualised so learners can decide in under 60 seconds without signing up. Features intelligent search and filters, a Gemini-powered recommendation chatbot, personalised knowledge-gap quiz, AI Insights editorial feed, progress tracking (saved, started, completed), and engagement-based trending rankings. Built on Next.js 16 (App Router, React 19) running on Cloudflare Workers via OpenNext, with Cloudflare D1, KV, and R2.

Curation PlatformEdTechAI ChatbotCloudflare
100+ Free Courses
20+ Providers

BytesCRM

AI-Powered CRM with Embeddable Chatbot Widget

A full-stack CRM for small businesses and freelancers combining an embeddable AI chatbot widget (single script tag, no developer needed) with a Kanban deal pipeline, contact management, autoresponder email sequences, and Stripe-powered tiered billing. Leads above the plan limit enter an overflow queue with a pay-per-lead unlock model. Built on React 18, Base44 serverless (Deno), with OpenRouter, Claude, and Gemini powering the chat AI.

Multi-tenantCRMAI ChatbotLead Capture
Free Tier
Pay-Per-Lead

Active Paw

Shopify Custom Theme: Pet Care Marketplace

A fully customised Shopify storefront for a UK pet care marketplace. Features collection-based navigation, review-led product confidence, free-shipping threshold prompts, and a care-first support UX optimised for pet owners browsing grooming, toys, and health essentials.

ShopifyE-commerceCustom ThemeUX
UK Market
Conversion-Focused

Pivot Parlor

Multi-tenant SaaS: AI-Generated Booking Sites for Barbershops

A multi-tenant SaaS platform where each barbershop or beauty salon gets its own AI-generated, fully branded booking site at pivotparlor.com/<shopSlug>: theme, copy, logo, hero image, and per-service photography generated per shop. Customers book as guests (no platform account required). Each shop has a dedicated AI chatbot grounded in its own live services, pricing, and promotions, answering only from real data. Shop owners manage staff, schedules, bookings, walk-in queues, waitlists, and revenue analytics via Stripe Connect. Built on Remix (Cloudflare Pages BFF) + Hono API (Cloudflare Workers), with D1, KV, R2, and Queues as a Turborepo monorepo.

Multi-tenant SaaSAI Site GenerationCloudflare EdgeGuest Booking
£100 Setup
10% Commission

Shikaku: Puzzle Quest

Competitive Logic Puzzle PWA: 100 Levels, Real-Time ELO Battles

A competitive browser-based PWA built with zero frameworks: vanilla HTML5, CSS3, and JavaScript with HTML5 Canvas dual-layer rendering and procedurally synthesized audio (no audio files to download). Features 100 hand-crafted Shikaku puzzles across 10 themed worlds with 3-star scoring, and a Battle Arena with ELO-rated 1v1 matchmaking, live opponent progress bars, fair-play puzzle rotation, and server-side solution validation via SHA-256 hashing to prevent cheating. Progress syncs across devices via Cloudflare KV; offline-first via LocalStorage with cloud migration on login. Deployed on Cloudflare Pages with serverless Workers API.

No FrameworkReal-time MultiplayerELO MatchmakingPWA Offline
100 Levels
1v1 ELO Battles

Oak Hill Park Cafe

Real-Time Booking Platform: Family Cafe and Soft Play

Full-stack platform for a Barnet family cafe: dependency-free static front end paired with a serverless booking system built on Cloudflare Pages Functions and D1. Features real-time party slot availability, a soft-hold workflow that prevents double-bookings at the database level, an owner admin panel, bot-protected enquiry forms with Resend email integration, and GDPR cookie consent. Pitched as a modernisation of their outdated site, timed around their new 5-star Food Hygiene Rating.

Booking SystemCloudflare D1ServerlessClient Proposal
5-Star Hygiene
No SaaS Dependency

Toolkit

Skills & Technologies

AI-assisted delivery combined with product and operations expertise: the practical toolkit for directing, reviewing and shipping real applications.

AI-Assisted Technical Toolkit

AI-Assisted Development Stack

Next.js / React project directionTypeScript-based app workflowsNode.js / API feature planningREST API integration conceptsTailwind CSS UI iterationAI-assisted debugging and review

Data & Storage

SQLiteDatabase DesignCloud Functions (Cloudflare Workers)

Platforms

Shopify (Custom Themes)GitHubCloudflareGoogle Cloud Platform (learning)

AI & Tooling

Generative AI IntegrationStructured PromptingClaudeChatGPTGitHub CopilotAI Governance (ISO 42001)Prompt Injection Awareness

Professional Skills

Product Development

Requirements DefinitionPRD DevelopmentWorkflow DesignPrompt EngineeringMVP ScopingFeature Prioritisation

Responsible AI

Input ValidationHuman Approval GatesOutput VerificationContext EngineeringHallucination Awareness

Operations & Commercial

E-commerce OperationsCRM SystemsCustomer RelationsRemote Team ManagementROI AssessmentProcess Improvement

Currently Learning

Google Cloud Fundamentals: Core Infrastructure

Responsible AI

How I Build Safe Applications

AI is a powerful tool. I use it with controls, not blind trust.

Input Validation & Data Quality

Validate format, type, and length on all inputs. Treat external content as untrusted by default. Reject unexpected or malformed inputs before they reach any AI or backend system.

Active Paw ghost-card quality gates · BytesCRM contact imports

Prompt Injection Awareness

Separate trusted system instructions from user-supplied content. Restrict what AI can access. Test adversarial inputs during development, not just happy-path scenarios.

PawScout LLM quality gates · BytesAI Learn chatbot and AI article generation

Human Approval Gates

AI-generated content requires human review before publishing. Consequential actions (releasing features, publishing pages, triggering automations) require explicit human sign-off.

Pivot Parlor: staff reviews AI pages before going live

Output Verification

Check AI outputs against requirements. Cross-check factual claims. Confident-sounding output is not a substitute for verification; hallucination is always a risk, treated as such.

PRD critique workflow · Case study content review

Permission Controls & Guardrails

Least-privilege access throughout. AI systems have no more authority than needed for their task. Structured outputs, rate limits, and error handling reduce blast radius from failures.

BytesCRM role-based access (User, Admin, Developer) · Stripe environment gating

Context Engineering

Maintain project purpose, requirements, decisions and task lists across sessions. Prevent context degradation from compounding errors. Compact and restart with a handover document when context limits approach.

This portfolio, built and maintained using this exact practice

Managing AI Governance in Organizations with ISO 42001

Alison · CPD Certified · April 2026

“Responsible AI is engineering discipline, not a compliance checkbox.”

Learning

Education & Certifications

Education

Certifications

View all 19 certifications on LinkedIn

Career

Work Experience

2023 – Present

AI-Assisted Application Developer (Independent Projects)

Self-directed · London, UK

  • Developed 6 live applications across SaaS, e-commerce, EdTech and games using a human-led, AI-assisted development workflow
  • Responsible for all product decisions: research, requirements, workflow design, architecture, testing, iteration and release
  • Applied responsible-AI principles throughout: input validation, human approval gates, output verification, and permission controls
  • Used structured context engineering and version control to maintain continuity across development sessions
July 2019 – December 2022

Remote Operations Manager / E-Commerce Specialist

Bytes Enterprises Ltd · London

  • Managed all customer communications and order operations remotely for a high-volume e-commerce business
  • Coordinated with 10+ suppliers and logistics partners to maintain seamless cross-organisational operations
  • Analysed sales data to streamline processes, reducing order-processing time by 20% while maintaining 99% inventory accuracy
  • Maintained detailed digital records using cloud-based tools, demonstrating strong digital systems proficiency
April 2015 – July 2019

Night Receptionist

Radisson Edwardian Hotels – The Grafton Hotel · London

  • Served as primary point of contact for guest needs during night shifts, handling check-ins, payments, and issue resolution with precision
  • Managed front-desk operations independently, demonstrating reliability and decision-making without direct supervision
  • Coordinated with multiple departments (housekeeping, maintenance, management) to resolve issues promptly and maintain guest satisfaction
April 2010 – March 2015

Central Reservations Agent

Radisson Edwardian Hotels – Central Reservations Office · London

  • Processed complex travel reservations using booking systems with 100% accuracy on client details, dates, and special requests
  • Suggested additional services to enhance guest experiences, demonstrating ability to add value and tailor recommendations
  • Trained new staff on reservation procedures and customer-service standards, showing clear communication and coaching ability

Connect

Let's Work Together

Have a project in mind or want to discuss how digital transformation can help your business? I am open to freelance opportunities, collaborations, and full-time roles.

Email[email protected]
Phone07429 578266
LinkedInlinkedin.com/in/fbmacagba

Based in London, UK

North London · Strong ties to local community · Flexible for on-site, remote, and hybrid roles