A software product company

Software that runs real businesses.

We build and operate our own products — and we rebuild how traditional companies work: digitizing, automating, and integrating AI where it measurably saves time and money.

16+
Products and platforms shipped
9
Industries served
1M+
Downloads on a single product
5
Languages, one site

Where we have shipped

  • Fintech
  • Hospitality
  • Government
  • Medicine
  • eCommerce
  • Manufacturing
  • Education
  • Tourism
  • Real Estate

Our products

Products we own and operate.

A product company first. We run our own platforms in production — and the patterns they prove are the ones we bring to every client system.

AI customer support

KulpunAi

An AI customer-support platform that handles incoming messages across WhatsApp, Telegram, Instagram, and the web — resolving most on its own and handing the rest to a human with full context. Meta-certified, and it runs on the client's own servers when security requires it.

80%of incoming messages resolved without a human, in production
4+
channels: WhatsApp, Instagram, Telegram, web
Meta
certified for WhatsApp & Instagram
2
deployments: cloud or on-premise
Visit

Manufacturing operating system

Garment Factory OS

One operating system for an entire garment manufacturer: the production floor, an employee app, a B2B wholesale portal, a consumer store, and an AI assistant leadership uses to query the whole company — all on a single data model.

Weekly planning & coordination hours
Before — ≈200 h/week
After — under 20 h/week
−90%
planning and coordination hours
5
surfaces on one data model
~200
hours/week of planning, before

Also in the line

  • Yurtah.AIoperating system for residential developers, 800+ units under management
  • Al-FajrAI-guided companion for Muslim daily life
  • Balam.AIMontessori parenting app with an AI teacher
  • MalSat.AIAI-powered livestock and meat marketplace

Selected work

Built for clients. Measured in their numbers.

A few engagements we can talk about — chosen because they shipped, they are recent, and the results are the client's, not ours.

Fintech

Mbulak

A microcredit app where the whole loan lifecycle lives on the phone — application, scoring, and disbursement without a branch visit.

1M+

downloads on Google Play

Hospitality

Karven Four Seasons

A resort operating system for a year-round resort on Lake Issyk-Kul — booking engine, guest apps, in-room tablets, staff operations, PMS dashboard, and a WhatsApp concierge on one data model.

7

surfaces, one data model

Public sector

AlaArchaPark.com

A super-app for Ala-Archa National Park: QR entry passes, accommodation and gondola booking, a tour marketplace, offline trail maps, an AI concierge, and a GPS-based SOS line to park rangers.

QR → SOS

the whole park visit, one app

Tourism

IssykKul.kg

A trilingual portal for the Issyk-Kul region — the lake, its resorts, and its season, presented with the calm of an official gateway and the polish of a consumer product.

3

languages, one regional portal

Hospitality

Kekeev

An end-to-end delivery system for a growing halal restaurant chain — ordering, dispatch, and operations owned by the brand, not an aggregator.

0

aggregator fees on own orders

Marketplaces

Kenguroo

An on-demand delivery marketplace built to compete with an international incumbent — grown into a real operating business, then sold.

Acquired

a complete build-to-exit

Research notes

What we learn when software meets a real business.

We write up engagements the way field researchers do — a question, a method, findings, and what they imply. Every number below comes from a shipped system.

RN–01Manufacturing · Operations

Removing coordination work from a garment factory

Abstract

A large garment manufacturer was spending roughly 200 hours a week on planning and task tracking spread across spreadsheets and chat. We asked how much of that coordination survives when every surface of the business shares a single data model.

Method

One operating system rolled out across five surfaces — the production floor, an employee app, a B2B wholesale portal, a consumer store, and a retrieval-based AI assistant for leadership — replacing the disconnected tools rather than integrating them.

Findings

−90%
planning & tracking hours
5
surfaces, one data model
1
source of truth
Fig. 1 · Weekly planning hours
Before — ≈200 h
After — under 20 h

Implication

Most coordination work in a traditional factory is not work — it is the cost of disconnected systems. Remove the disconnection, and the hours go with it.

RN–02Service operations · AI

How far can AI take frontline customer support?

Abstract

Support teams answer the same questions across four or five channels, slowly and inconsistently. Running our own platform in production let us measure how much of the incoming stream a multi-model AI layer resolves without a human.

Method

KulpunAi in production: multiple frontier models behind a routing gateway, unified across WhatsApp, Telegram, Instagram, and web chat, with CRM context and a human handoff that preserves the full conversation.

Findings

80%
auto-resolved, up to
4
channels, one brain
100%
of handoffs carry context

Fig. 2 · Share resolved by AI

80%of incoming messages, without a human

Implication

The ceiling on AI support is not model quality — it is context. Given the business's own data and a clean handoff, most volume never needs a human.

RN–03Consumer fintech · Distribution

In emerging-market lending, distribution is the product

Abstract

Microcredit in most emerging markets runs on branches, paperwork, and slow manual decisions. We asked what happens to reach and volume when the entire loan lifecycle — application, scoring, disbursement, repayment — moves onto the phone.

Method

A mobile-first lending app with an automated application and scoring flow, built to national scale from day one, so a borrower goes from install to credit decision without visiting a branch.

Findings

1M+
downloads on Google Play
0
branch visits required
National
operating scale
Fig. 3 · Branch visits per loan
Before — at least one
After — zero

Implication

Where desktop banking never arrived, the phone is not a channel. It is the branch, the loan officer, and the ledger at once.

How we work

Traditional problems, modern operating systems.

Most businesses do not need more software. They need fewer systems that disagree with each other.

  1. 01

    Diagnose

    We map how the business actually runs — the spreadsheets, group chats, and workarounds — and find where hours and money quietly leak.

  2. 02

    Digitize

    Paper, memory, and tribal knowledge become structured data on one model, so every team finally sees the same truth.

  3. 03

    Automate

    The repetitive work — planning, dispatch, reporting, follow-ups — moves into the system and stops consuming people.

  4. 04

    Integrate AI

    AI where it measurably pays: answering customers, drafting documents, and letting leadership query the whole company in plain language.

What typically comes back

From shipped systems — not projections.

Planning & coordination hours· Garment Factory OS

−90%

Customer messages resolved by AI· KulpunAi

80%

Contract review time· Yurtah.AI

−70%

The result is rarely dramatic on day one — and reliably dramatic by month three: fewer hours lost to coordination, fewer errors, and a business that can finally see itself clearly.

The Studio OS

The engagement is a product, too.

You don't hire us into email threads and PDF attachments. Every client works inside a private portal — the same operating-system discipline we sell, applied to working with us.

Request a private briefing
  1. 01

    A proposal, not a PDF

    Your commercial proposal is delivered inside the portal — scoped, priced, versioned, and approved in one place. Nothing gets lost in an inbox.

  2. 02

    Requirements with numbers

    Product requirements arrive with unit economics attached: what it costs, what it replaces, and when it pays for itself.

  3. 03

    A build you can watch

    Roadmap, decisions, and working software appear in the portal as they happen — not in a monthly status call.

  4. 04

    Results that report themselves

    After launch, the system publishes its own metrics to your portal — the same numbers we later write up as research.

Global by default

One studio. Five languages. Any market.

We let the work speak and keep the volume low. We ship in English, Kyrgyz, Russian, Chinese, and Arabic — and we have delivered in fintech, hospitality, government, medicine, and eCommerce.

Partnerships & investment

We are open to global partnerships, distribution alliances, and conversations with investors who want exposure to working software solving real, revenue-relevant problems. We are curious, and we are not in a hurry.

Contact

Request a private briefing.

Tell us what should change in your business. Within 24 hours you get a concrete assessment from the founding team — not a sales script.

What do you need?

Confidential by default. Your message goes straight to the people who build.