Client work

Client case studies

Work we have delivered for clients, with the results and the stack behind each project. We build first and theorize later.

Results at a glance

$3K+ → <$500MAUA's monthly hosting after the move to a leaner stack
25 → 4code repositories for MAUA, live in Samoa and Vanuatu
90%of routine retail inquiries resolved by AI agents during campaigns
2-6 weekson average to put the first automated workflows live
Venture building

Looking for our ventures?

We also co-found digital products: Second Brains OS, AskLawyer.vn, LOKO.travel and Sourcing.vn.

See our ventures →
01 · eCommerce & ERP
MAUA logo

MAUA Superapp Platform

We consolidated 25 code repositories into 4 for MAUA, the Pacific's leading e-commerce provider. Live in Samoa and Vanuatu, it now runs on a leaner Bun stack, and hosting fell from over $3,000 a month across AWS, GCP and DigitalOcean to under $500.

Overview and stack
Challenge
Operations in Samoa and Vanuatu ran separately, inventory synced slowly, and the AWS bill was very high.
Strategy
Consolidate everything into one monorepo and a single Bun monolith built for high concurrency.
Stack
Bun, Hono, Next.js/Remix, PostgreSQL, PostGIS, Mosquitto MQTT.
Read the case study → maua.app ↗
02 · Infra strategy

Cloud to On-Prem

A migration guide for moving heavy data models and compute pipelines off public cloud services and back onto self-hosted hardware clusters, with only a low-cost edge proxy left in front for public traffic.

Overview and stack
Challenge
Cloud bills and DevOps overhead kept growing.
Strategy
Move to self-hosted local clusters behind zero-trust Cloudflare tunnels.
Stack
Cloudflare Tunnels (cloudflared), Raspberry Pi clusters, Mac Mini nodes, Gitea.
Read the case study →
03 · AI workflows

Voice & Chat AI

We deployed AI agents that route customers within fixed rules during high-volume retail campaigns. They resolved 90% of routine inquiries on their own. For complicated shipping cases, they gathered the details first and handed the customer to a person with the full context.

Overview and stack
Challenge
Ticket volume spiked during promotional sales and support fell behind.
Strategy
Low-latency AI agents that pull in local context and hand off to a person when needed.
Stack
Hono API, Groq LLM endpoint orchestration, contextual prompt injection.
Related: AI customer service for F&B →