Meet the team (all two of us)

Keelwork is a small studio in Da Nang, Vietnam. One of us has been building software since 2017 and writes all the code. The other one is new.

  • FounderHuynh Ngoc Hai, Principal Engineer, coding since 2017
  • CollaboratorIn onboarding, naps a lot
  • OfficeDa Nang, Vietnam
  • ClientsAnywhere with Wi-Fi
Keelwork's founder in a straw hat, holding their baby son against an orange wallthe senior engineerwrites the codethe collaboratorstill in onboarding

Why Keelwork exists

Most founders who have hired an outside team have a story about one that vanished. The developer who stopped replying in month three. The demo that worked, and the product nobody finished.

Keelwork exists so that doesn't happen to you. A keel is the part of a boat you never see. It sits below the waterline and keeps the boat upright and on course, whatever the weather. That's the job.

You hear from us every week, including the weeks when the news is bad. Your code lives in your repository from the first day, so your product never depends on us staying around.

It's a business, and I won't pretend otherwise. But the only version worth running is the one where clients come back.

What we believe

A demo takes a week. Keeping it right for a year is the actual job.

The senior engineer. The collaborator declined to comment.

Been there, shipped that

Enterprise systems serving millions of users, in fintech, edtech, logistics, healthcare and more, built in teams of 30+ engineers. The client names stay under NDA.

  • FinTechPayments, lending, accounts
  • EdTechLearning platforms
  • LogisticsShipping and operations
  • HealthcareClinical and patient systems
  • E-commerceStores and marketplaces
  • B2B SaaSProducts for business teams

AI systems we have built

Three kinds of work, each taken past the demo and into production.

  1. Answers from documents

    RAG assistants over policies, contracts and manuals, where every answer cites the passage it came from.

  2. Documents into data

    LLM extraction that turns invoices, forms and shipping papers into structured records a system can use.

  3. AI inside existing products

    AI features added to web and mobile products already in use, with eval sets and a cost cap per request.

Hands-on AI experience

The parts of an AI build that decide whether it survives real users.

  • RetrievalChunking, hybrid search, reranking
  • Vector storepgvector on Postgres
  • ModelsOpenAI, Anthropic
  • EvalsTest sets and regression runs
  • AgentsTool calling with approval steps
  • Cost and latencyCaching, routing, streaming
  • SafetyGuardrails and refusals

Small on purpose

One senior engineer leads every project and writes its code. That is the point, not a stage before growing.

When a project needs more hands, trusted collaborators join for it. You meet each one before they start, and the senior engineer stays the person you talk to.

No account managers and no handoffs to a junior team. The engineer on the consult is the engineer on the code.

How a project runs