Full-stack SaaS developer with 13+ years of experience building multi-tenant business applications,
workflow automation systems, and web/mobile platforms using Laravel, Node.js, Vue.js, React,
and AI-assisted development.
SaaS applications and software products designed around real business workflows, multi-role access, and cross-platform delivery.
Experience
Client & Professional Work
Production systems delivered for international clients across analytics, insurance, logistics, and business operations.
DevOps / Fullstack Developer
Building and maintaining a multi-tenant analytics platform for Amazon sellers, integrating three ad ecosystems into a single dashboard with AI-generated insights and automated data pipelines.
Business Data Analytics Dashboard: consolidates Amazon SP-API (business reports, FBA inventory, customer repeaters), Amazon Ads API (sponsored products/brands/display, DSP, AMC), and Meta Ads into a single Laravel + Vue 3 + Vuetify interface. Nightly scheduler pulls per-company reports; includes AI-generated insights with EN/JA translation and an Automatic Report Import page with auto-retry on failed fetches.
AI
Production AI Feature
AI Integration Experience
I built a production-ready AI Integration that generates Product Insights for Amazon sellers. The feature allows users to create AI-powered analysis for each product based on sales, advertising, inventory, and Amazon Business Report data. The backend collects the relevant metrics, creates a structured prompt, sends it to an AI API, and stores the generated response so users can view the insights instantly without regenerating them each time.
500k
Amazon Records Processed Nightly
$240/mo
AI API Cost (down from $1,200)
50+
Amazon Seller Accounts
Learn More
I designed and implemented this entire AI workflow, from collecting Amazon data and building optimized prompts to caching AI responses and displaying insights in the application.
How the AI insight feature works
1. User opens insightsThe seller clicks the AI insights button for a product or report period.
→
2. Check saved result firstThe system checks the database for an existing insight so users do not wait or pay for the same analysis twice.
→
3. Show or generateIf found, it displays instantly. If not found, the backend creates a new AI request.
↓
Generate pathCollects sales, ads, inventory, and Amazon report data from Amazon SP-API and Amazon Ads API.
Prompt pathUses configurable Prompt Engineering rules, then sends the structured request to Claude AI.
Store resultSaves the AI evaluation, recommendations, raw response, token usage, and status in the database.
Display insightShows Product Insights, recommendations, KPI comparisons, and optional EN/JA translation in the frontend.
In simple terms: I built the feature that turns Amazon seller data into useful business recommendations. It first checks if an insight already exists, shows it instantly when available, and only asks AI to generate a new one when needed. This makes the feature faster, cheaper to run, and more reliable for real users.
I also built a configurable prompt management system, allowing prompts to be updated without changing the application code. The AI analyzes product performance, advertising metrics, profitability, inventory levels, and business trends to generate actionable recommendations for Amazon sellers. My responsibilities included designing the backend integration, preparing and aggregating data from Amazon SP-API and Amazon Ads API, constructing optimized prompts through Prompt Engineering, handling Claude AI API requests, storing AI responses, and integrating the results seamlessly into the frontend.
Employer Testimonial
This employer review is visible on my OnlineJobs.ph profile when viewed from an employer account. The employer name and logo are blurred for privacy.
LaravelVue.jsInertia.jsMySQLUbuntuDigitalOceanAI IntegrationClaude AIPrompt EngineeringAmazon SP-APIAmazon Ads API
Fullstack Web Developer
Built and maintained multiple production-grade client systems across insurance, logistics, fundraising, and e-commerce domains.
Insurance Management Platform: handles certificates of insurance, contracts, claims, renewals, invoicing, and rate calculations. JWT-authenticated REST API consumed by a Vue.js frontend. PDF generation, role-based permissions via Spatie, AWS S3 file storage.
Inventory / Campaign / Order Management: multi-tenant Laravel/Inertia app for promotional campaigns, candidate recipients, inventory bundles, and shipping fulfillment. Deployed on AWS with S3-backed assets and Pusher real-time notifications.
Marlon really is a strong backend focused dev with specialization in Laravel and Drupal development. He's a true workhorse capable of pulling the heaviest and hardest workloads.
This testimonial can be seen on the company website.
Fullstack development across a mixed tech stack for a German-based client, working on web applications using modern JS and PHP frameworks alongside Azure serverless functions.
Android applications currently in testing and preparing for public release. APKs marked “In Review” are staging/beta builds shared with early testers for feedback ahead of their Google Play launch.
Swipe for more apps
My Internal Operations
How I Maintain My Applications
I use automation to collect production errors from my apps and servers into one centralized maintenance workflow. Errors are grouped and analyzed with AI, converted into structured Trello issues through n8n, then investigated, fixed, tested, deployed, and monitored.
My Apps & Their Servers
Circlst
RentaDiary
CarFleex
TagasReport
BalanseMo
Error Sources Monitored
Key services and platforms are continuously monitored for runtime errors,
server failures, Android crashes, ANRs, and backend issues across my applications.
◇
5Total Sources
⌁
5Active Monitoring
▤
5Synced Reports
Active
Supabase Logs
Database, backend, authentication, and API errors from Circlst.
View details →
Collected from Circlst’s Supabase backend to investigate database, authentication, and API failures.
Active
Apache Logs
HTTP request failures and production server exceptions.
View details →
Server-side error context helps trace failed requests to the affected application and code.
Tracked
Node.js / PM2
Runtime exceptions, process crashes, unhandled rejections, and API errors.
View details →
Process and application logs help investigate crashes, rejected promises, and service failures.
Tracked
Google Play Crashes
Android crash reports, ANRs, affected versions, and device information.
View details →
Android vitals reports provide crash and ANR context by app version and device.
Synced
Firebase Crashlytics
Mobile crash events, stack traces, affected sessions, and app diagnostics.
View details →
Crashlytics reports provide diagnostic context for apps with the Crashlytics SDK installed.
I centralize crash logs and production errors from my applications and servers,
analyze recurring issues with AI, create structured reports, investigate the affected code,
and monitor each fix after deployment.
I built CrashWatch that automatically collects and centralizes crashes, server errors, application logs, and monitoring context.
Apache logs
Node.js / PM2 errors
Supabase logs (Circlst)
Google Play crashes & ANRs
Firebase Crashlytics
2
Analyze
Analyze Errors
AI-assisted analysis using Gemini Flash to group recurring failures, inspect stack traces, identify affected apps, and suggest likely root causes.
Group recurring errors
Inspect stack traces
Identify affected code
Prioritize by severity
Gemini FlashAI Analysis
3
Report
Create an Issue Report
n8n automates the workflow, generates a structured issue report from the findings, and creates a Trello issue for tracking and resolution.
Structured error report
Automatically create Trello issue
Attach logs and context
Set priority and status
n8nTrello
4
Fix
Investigate & Fix
I review the Trello issue and use AI-assisted development tools such as Codex and Claude to investigate the codebase, identify the root cause, and implement the fix.
Review code and production logs
Use AI-assisted suggestions
Implement and test the fix
Update the Trello issue
CodexClaude
5
Monitor
Review & Monitor
I review the changes, run tests, deploy through CI/CD, and monitor the application to verify that the issue is resolved.
Deploy via CI/CD
Monitor for new errors
Confirm the fix is effective
Close or reopen the issue
CI/CD
↻Continuous improvementEach resolved issue feeds back into a more reliable production system.
A screenshot of my real profile, not a badge I made
95/100ID PROOF SCORE
Government ID verified · Timeproof enabled
Ask me for my OnlineJobs.ph profile link, I'll send it right away.
Shipped
BalanseMo, approved for Google Play
Granted production access, live for real users on Google Play.
Product builder
I design the feature set, not just the code
I don't start from a competitor's feature list. I learn the actual process, how the
work already gets done, then put myself in the user's place and build
what that person needs to finish the job.
4
Products designed this way
Solo
Research, design, build, and ship
End to end
Whole workflow, not one step
I benchmarked against Google Play Apps only after shipping a stable version, and
found my products covered workflows the established apps left out.
Understanding the process turned out to matter more than matching a feature list.
Production access granted
RentaDiary, approved for Google Play production
Google approved RentaDiary for production access, and the app is now available to users on Google Play.
Direct request
Request for Product Demonstration
A water utility requested a live demonstration of TagasReport before approval.
Monetization approved
BalanseMo is ready to show AdMob ads
Google completed the app review and lifted the ad-serving limits for the production Android release.
How I work now
AI-assisted development
AI is part of my daily workflow: Claude Code, GitHub Copilot, and ChatGPT,
on a multi-tenant analytics platform and my own products under DizonLabs.
Scoped sessions: separate branches and worktrees so parallel AI work can't collide
Every change reviewed: I read the diff before it merges; nothing ships unread
Verified, not assumed: real builds and tests prove it works before I call it done
It removes the boilerplate so my time goes to architecture, data modeling, and the
business logic that actually needs judgment.
Beyond the code
Why I still love this
Thirteen years in and the good part hasn't changed: someone describes a messy, manual
process, and a few weeks later it just… runs by itself.
That's the whole job to me. Not frameworks, giving people back their time.
On GitHub
The work, day by day
2,204Contributions past year
263Active days of 365
33Longest streak
20 public repos: PHP, Vue, TypeScript, Node, Dart, and Python
Contributing since 2015, still shipping this month
TagasReport
