Backend Engineer · Bangkok, TH · GMT+7

Aom KhamPhaung

Ships multi-tenant SaaS end-to-end — ambiguous requirement to production — for enterprise clients across Japan and Southeast Asia.

Scroll
Now
Backend Engineer, Coosy — Japan, remote
Since
2022 · 3+ years shipping production
Core
Laravel · FilamentPHP · AWS · MySQL
Status
Open to relocate/remote roles

What Iactually do

Three things, no filler
Ownership

Own it end-to-end

Ambiguous business requirement → technical spec → production deploy → post-launch monitoring. No handoff gaps, no waiting to be told.

Architecture

Build for scale

Multi-tenancy, RBAC, relational schema design, Redis and Cloudflare caching, Laravel queues, zero-downtime releases on AWS.

AI-native

Ship with AI in the loop

LLM APIs shipped inside client products, and AI coding tools in the daily workflow — faster output without dropping review standards.

Trackrecord

2022 — present
2024 — PresentCoosy
Japan · Remote

Backend Engineer

Multi-tenant SaaS CMS for enterprise clients

  • Architected a multi-tenant SaaS CMS in FilamentPHP and Laravel for 10+ enterprise clients — RBAC, REST integrations, automated provisioning. Onboarding time down 40%.
  • Took full product ownership: ambiguous requirements into technical specs, through deployment and post-launch monitoring.
  • Integrated LLM chatbot APIs into client applications for context-aware automation.
  • Established root-cause analysis and structured debugging workflows across environments. Downtime down 25%.
LaravelFilamentPHPAWSDockerRedisCI/CD
2022 — 2024ILBC
Myanmar · Onsite

Junior Software Developer

APIs, integrations and performance work

  • Designed and optimized relational schemas that held up as enterprise workloads grew.
  • Led REST API development and third-party integrations — Google APIs, OAuth, SMS gateways.
  • Shipped payment gateway integrations with transaction reliability and compliance as the bar.
  • Cut page load times and raised capacity with Redis and Cloudflare caching strategies.
PHPMySQLRedisCloudflareGit

Selectedwork

Two products, one platform
Flagship product · multi-tenantClient work · pre-launch

AI Chatbot SaaS Platform

Backend engineer — architecture, APIs, background pipelines

A tenant-isolated chatbot service that answers from each customer's own knowledge base. Tenants ingest their material — uploads, documents, crawled pages — and the platform chunks and categorises it, then serves grounded answers through an embeddable widget with live streaming.

  • Knowledge ingestion pipeline: file, CSV, document and URL crawling, each on its own queued job so a large import never blocks the request cycle.
  • Chunking and category redistribution with recrawl support, so a tenant's corpus stays current without a full reimport.
  • Realtime widget messaging over WebSockets, with generation success and failure broadcast as events rather than polled.
  • Plan and usage metering per tenant, subscriptions, invitations and role-based access across tenant boundaries.
  • Media and document storage on S3-compatible object storage, with dashboard widgets prefetched and cached to keep the admin panel responsive.
Laravel 12FilamentPHP 4Reverb / WebSocketsQueuesLLM APIsAWS S3ReactDocker
ClientAPIWorkersData
Embeddable widget

A script tag a tenant drops onto their own site. Domain-restricted per tenant, so a stolen embed key will not answer from someone else's knowledge base.

ReactViteDomain allowlist
Companion product · same platformClient work · pre-launch

Multi-Tenant Headless CMS

Backend engineer — content modelling, tenancy, integrations

The content side of the same platform. Each organisation manages its own sites, custom post types, taxonomies and field groups, then publishes through an API the chatbot and front-end clients both consume. Sign-in runs through a shared OAuth portal, so one account reaches every service.

  • Content modelling built for editors: custom post types, taxonomies, terms, and repeatable field groups rather than fixed schemas.
  • Organisation → site → user tenancy with invitations and per-site roles, so agencies manage many client sites from one account.
  • Single sign-on across all services via an OAuth portal, with impersonation for support staff and an audit trail behind it.
  • SEO metadata, previews, rich-text editing, PDF export and webhooks for downstream systems.
  • Billing surfaces — plans, invoices and line items — shared with the platform's other services.
Laravel 12FilamentPHP 4InertiaPassport / OAuthMySQLAWS S3Docker
ClientAuthAPIData
Editor panel

Where editors model and write content: custom post types, taxonomies, terms and repeatable field groups, rather than a schema fixed at build time.

FilamentPHP 4TiptapInertia
A note on these

Both are client products that have not shipped publicly, so they are described at architecture level only — no screenshots, client names or source. Happy to walk through the design decisions in a conversation.

Technicalwriting

Published · AWS Builder Center
October 2026AWS Builder Center

What If Your AWS Architecture Could Review Itself?

A look at the AWS Well-Architected Agent in public preview: it reads resource configuration, utilisation metrics and application topology across 65+ services, then ranks findings against the priorities you actually stated — reliability, cost, performance, security. The part that matters is the output: IaC changes, CLI commands and SSM runbooks, which turns architecture review from a periodic snapshot into a feedback loop that runs continuously.

Well-Architected AgentArchitecture reviewGenAI for infra
September 2026AWS Builder Center

Stop Writing Glue Code: Building Event-Driven Pipelines with Amazon EventBridge Pipes

Most service-to-service integration is a Lambda that filters, reshapes and forwards — code you own, test and page on. EventBridge Pipes does filtering, enrichment and transformation as configuration on a point-to-point connection instead. Pipes isn’t replacing application logic; it removes the plumbing around it, so the code you keep is the part that carries business value.

EventBridge PipesEvent-drivenLambdaSQS
September 2026AWS Builder Center

Your S3 Bucket Knows More Than You Think: Turning Object Storage into an AI Context Layer

S3 Metadata and S3 Annotations expose object context as Apache Iceberg tables you can query with Athena — which means business context can live next to the object instead of in a side metadata database you have to keep in sync. The argument for agentic workloads: agents discover and filter the right data themselves, because the data carries its own context.

S3 MetadataS3 AnnotationsAthenaAgentic AI
A note on these

Written for the AWS Builder Center under my own byline — the same work as writing a technical spec, aimed at readers who have to make a decision at the end of it.

All articles on AWS Builder Center

Thestack

Honest about depth

Backend

PHP · LaravelFilamentPHPNode.jsGolang · learningPython · learning

Frontend

ReactNext.jsVueTypeScript · familiarTanStack · familiar

Data

MySQLPostgreSQLRedisMongoDB

Cloud & DevOps

AWSDockerGit · GitHubGitHub ActionsCI/CDGCP · familiar

AI & Tooling

LLM API integrationAI coding assistantsPrompt engineering

Practices

SOLIDREST designMulti-tenancyRBACSpec writingAgile / SCRUM

Filled = production depth · Outline = learning or familiar

WorkingAI-native

AI is in the loop every day, but it does not make the calls that matter. The split is deliberate.

  • AI draftsFirst-pass specs, scaffolding, test stubs, refactor sweeps, unfamiliar-API exploration.
  • I decideArchitecture, schema design, tenancy boundaries, and what ships to production.
  • ResultMore surface covered per sprint with review standards intact — bug reports still went down 30%.

Education

BSc Computer ScienceUniversity of the PeopleExpected Nov 2026
Associate Degree, Computer ScienceUniversity of the PeopleJan 2026
Computer Applications & ITArizona State UniversityJan 2024

Hiring backendor full-stack?

I reply within a day. Remote-ready from Bangkok, used to working across Japan and Southeast Asia time zones.

Bangkok, TH · GMT+7Open to relocate/remote