CUSTOM AI

We design, build and run bespoke AI systems

for firms whose work doesn't fit off-the-shelf software

REAL WORK

Engineered alongside your team,

on your data

EXCELTECH

Applied AI lab,

Singapore

NCSNTT DATADBSHCLTechJ.P. MorganSingapore AirlinesLGTNCRCIMBLG CNSTech MahindraOatlyProtolabsDigitalOceanPalomarGuardant HealthRadiant LogisticsCharles River AssociatesUdemy
ExcelTech
SINGAPORECONTACT

What your firm knows, engineered into systems that act on it.

// Selected work

Built around
your firm's
real work.

Custom AI systems for enterprises and firms running large operations — banking, insurance, healthcare, logistics, retail, manufacturing, energy, F&B, real estate, government and more.

Wealth & family office

A Singapore multi-family office

Client onboarding, from first conversation to an open bank account.

Onboarding a new family meant weeks of forms and hand-offs between the office, corporate secretaries and banks. We rebuilt it as one tracked pipeline, with the team signing off at each gate.

  • An onboarding agent that interviews each client on background and source of wealth.
  • One tracked pipeline from incorporation to account opening — documents generated from verified fields, adverse-media screening on every related party.
  • A governed knowledge agent and document portal, deployed inside the client's own cloud.

Agents·Workflow automation·Document generation·Private cloud

Schematic of the wealth & family office system.

Interior design

A Singapore interior design and renovation firm

Show homeowners the room — and the budget — before the first meeting.

Homeowners arrive with a photo and two questions: what could this look like, and what will it cost? We built both answers into the firm's website.

  • Room restyling: upload a photo, pick one of the studio's styles, see your own space reimagined.
  • A quoting agent that interviews the homeowner and returns an indicative budget from the firm's own price ranges — never from the model.
  • Every conversation handed to the designers as a lead, with the brief attached.

Image generation·Conversational agents·Pricing logic

Schematic of the interior design system.

Solar energy

A Singapore solar energy company

From a postal code to a panel count.

Sizing a rooftop used to start with a site visit. Now it starts with a postal code — and anything uncertain goes to a person.

  • Postal code to the right building, with checks that catch a neighbour's roof before it reaches a quote.
  • Roof analysis face by face — pitch, orientation, usable area — with a panel count and a yearly energy estimate.
  • A review screen showing source and confidence, so an operator confirms or corrects in one click.

Geospatial·Computer vision·Human review

Schematic of the solar energy system.

Oil & gas trading

An oil & gas trading desk

Event context for an oil desk — without pretending to forecast.

Prediction markets reprice fast around sanctions, supply and weather. The desk wanted those moves beside its own data, every claim traceable, and no black box calling the price of oil.

  • Continuous collection of oil-relevant contracts, with fixed rules for which moves become events.
  • An evidence drawer with exact source links and cross-checked news — gaps are labelled, never filled in.
  • Findings reach the desk's timeline only after human review.

Market data·Event detection·Human review

Schematic of the oil & gas trading system.

Marine engineering

A marine engineering firm

From an RFQ in the inbox to a priced quote — without retyping a spec.

RFQs arrive as emails and attachments, and every one was re-keyed by hand into a quote. We built a pipeline that reads each RFQ, drafts the quote from the firm's own rates, and holds it in a review queue before anything is sent.

  • Every RFQ email and attachment parsed into structured line items — gaps flagged and held, never guessed.
  • Prices come from the firm's own rate sheets in code; the model only writes the covering email.
  • A review queue where the commercial team approves, edits or rejects each draft before it leaves.

Email agents·Document extraction·Pricing logic·Human review

Schematic of the marine engineering system.

// About

Applied AI,
engineered
on site.

ExcelTech is a Singapore-based applied AI lab. We design, build and run bespoke AI systems for firms whose work doesn't fit off-the-shelf software.

We work forward-deployed: our engineers sit with your team, learn the real workflow — the documents, the exceptions, the hand-offs — and ship systems that run on your data, inside your perimeter.

We move fast, cut through noise, and prize clarity, trust and technical excellence.

// How we build

In every engagement

Models draft. Code decides.

Figures, filings and documents come out of deterministic code. Models handle language — never the final number.

Your data stays yours.

We deploy inside your cloud or network, and route client data only to the providers you approve.

Uncertainty on screen.

Sources, confidence and missing data are shown, not smoothed over. Anything unsure goes to a person.

Engineers in the room.

Forward-deployed engineers work beside your team from the first workshop to go-live, and stay to run it.

// What we build

Systems built around
how your firm works.

Your workflow + Your data + Our engineers = Systems that run

Workflow & operations

Multi-step processes run end to end — intake, checks, documents, approvals and hand-offs — with an audit trail at every step.

Onboarding·KYC·Documents

Knowledge & retrieval

Governed answers over your own documents: access control, citations and audit, on the models you approve.

RAG·Access control·Audit

Client-facing AI

Tools your customers touch — visualisers, quoting agents, guided intake — that hand a clean brief to your team.

Visualise·Quote·Intake

Vision & data intelligence

Imagery, geospatial and market data turned into measured, reviewable outputs rather than confident guesses.

Imagery·Geospatial·Market data

Workflow & operations

// How we work

How we work

From first workshop to go-live

  1. 01 Discover

    A working session on the real workflow: the documents, the exceptions, the hand-offs.

  2. 02 Embed

    Our engineers sit with your team and map the real workflow, documents and edge cases.

  3. 03 Prototype

    A working system on your data early, shaped with the people who will use it.

  4. 04 Build

    Hardened for production: access control, an audit trail, and human review where it matters.

  5. 05 Deploy

    Deployed inside your perimeter, on the models and providers you approve.

  6. 06 Run

    Monitored, measured and supported by the engineers who built it.

  7. 07 Improve

    Tuned with you after go-live, as the work and the models change.

// Clients

Our
clients.

// Get in touch

Bring us
the hard
workflow.

Tell us about the workflow, the documents and the people involved. We'll come back with how we'd approach it.

recruit@exceltechcomputers.com
contact.sh0 / 4 fields

## ExcelTech contact form

Fill in the fields below. Items marked with * are required.

exceltech ~ % ./send_message.sh

⌘ / Ctrl + Enter