Senior engineers.
Systems that survive
production.
- Location
- Almaty · GMT+5
- Shipping since
- 2017
- Products live
- 14
- Domains
- Fintech · Enterprise · EdTech
- Status
- Taking work
Fast, built properly, and priced without the agency markup. You pay for senior engineers writing code, not for account managers and bench time. Eight weeks from signature to a working system.
users on systems
our engineers built
documents searched
in 500 ms
products shipped
and still running
building production
software since
Four things. All of them are running in production somewhere.
Each one is already running on a live system that our engineers own.
- AI
AI apps and integration
Agents and LLM features connected to your real backend. Sales agents in WhatsApp and Instagram. Search across your own documents. KYC and AML checks. Call scoring in Russian and Kazakh. Everything runs where your data already lives.
Claude / OpenAIRAGAgentsWhisper RU+KZLangfuse - WEB
Web platforms
Products that carry real load. One of ours runs a national exam platform at 1,000 requests per second. Another is the corporate information platform for a global logistics group.
Java / SpringNuxt 4ReactPostgreSQLKafka - MOB
Mobile, in Flutter
One codebase, both stores, from architecture to release. Offline sync, payments, push, native modules. One of our apps has passed 100,000 installs on Google Play.
FlutterDartBLoCApp StoreGoogle Play - OPS
DevOps and architecture
Kubernetes, Helm and GitLab CI/CD on AWS, Azure or your own servers. Releases without downtime, autoscaling, monitoring. We also take over infrastructure someone else left behind.
KubernetesHelmTerraformProxmoxObservability
Fourteen products. Nine you can open right now.
Every link below opens a live product. We check them before each deploy.
001Juz40National learning platform — Kazakhstan40,000+ users · 1,000 RPS
Exam season in Kazakhstan is unforgiving. The whole country studies in the same two months. We built Juz40 to hold 1,000 requests per second with more than 40,000 active users. The mobile app raised engagement by 65%.
Java / SpringFlutterPostgreSQLKubernetesOpen ↗ — opens in a new tab002EduserEdTech platform with an AI advisor100,000+ installs
A national EdTech platform built around iDos, an AI advisor that turns an exam score into realistic odds of winning a state grant. Nuxt on the web, one Flutter codebase for iOS and Android. Over 100,000 installs on Google Play and a live App Store release.
FlutterNuxt / Vue 3Java / Spring BootLLMElasticsearchKafkaPostgreSQLOpen ↗ — opens in a new tab003R.I.N.G.Corporate platform for a global logistics groupEnterprise · live
A corporate information and communications platform for the Kazakhstan arm of one of the world's largest freight forwarders. It runs on the client's own corporate domain.
Java / Spring BootAngularEnterprise SSO004SellioneAI sales agent — Instagram and WhatsAppAI agent in production
An AI sales rep that works inside Instagram and WhatsApp DMs. It reads what the buyer wants, shows the products, closes the order, then hands off payment and delivery.
LLMWhatsApp Business APIInstagram Graph APIOpen ↗ — opens in a new tab005AquaCRMVertical SaaS for car-wash networksLive in 5 cities
Car washes were running on paper journals. Now an attendant opens a job on a phone at the bay in three taps, and the owner sees revenue per location without driving to the office. Commercially live in five cities.
FlutterPHP / LaravelWeb back officePostgreSQLOpen ↗ — opens in a new tab006TakonDigital token platform for corporate spendFintech · mobile
Digital tokens in place of paper vouchers for corporate spend. Issue, redeem and reconcile, all from a phone.
FlutterPaymentsJavaOpen ↗ — opens in a new tab
More products
- Akadem.kzNational olympiads and teacher certification↗
- Ustaz.mediaEdTech platform with an AI teaching assistant↗
- Monolit GIDCorporate knowledge portal and Flutter app↗
- ILogistCargo logistics marketplace
- AllThingsNo-code online store builder
- GreenGoGeo-based marketplace for flowers and plants
- IQRAKazakh language learning app
Our engineers did not learn this on your budget.
Our engineers have built and led platforms in banking, telecom, marketplace search and global logistics. Every domain below is something they have run in production, under real load and real regulators.
Fintech and payments
Multi-acquiring with the country's largest banks. BNPL and split payments. Apple Pay and Google Pay. A consumer wallet with cross-border transfers. A lending stack with its own anti-fraud. AML checks driven by an LLM.
Banking
Corporate online banking, and a monolith-to-microservices migration inside a bank. A bad release there is a regulatory incident, not a rollback.
Telecom
Super-app backends for national mobile operators: microservices, in-app payments, integrations with banks and the enterprise service bus. Shipped on both stores.
Search at scale
A marketplace search and personalisation system over three billion documents. Response time 500 ms, with more than ten million index updates a day.
EdTech
Five platforms in production, including a national exam-prep system at 1,000 requests per second and an AI advisor with six figures of installs.
Logistics
A corporate platform for a global freight forwarder, and a cargo marketplace that connects shippers with carriers.
No juniors. Ever.
The people who sit in your process interviews write the code and stay on call after launch.
- 01
Every engineer has held a CTO or Tech-Lead title
Across fintech, EdTech, agritech and systems integration. Our people have set architecture and built the teams that shipped it.
- 02
One of us has trained 400 Java developers
We have taught this work as well as done it. If we cannot explain a decision in plain language, we do not ship it.
- 03
Enterprise scale, boutique speed
Most teams are one or the other. The engineers on your project have run platforms where a bad release is a regulatory incident, and they can still put a working AI system in front of your users in eight weeks.
- 04
Three working languages
We run projects in English, Russian and Kazakh. For AI work here that matters more than it sounds. We have built speech systems that handle Kazakh, including speech that switches between Kazakh and Russian mid-sentence.
Four steps. Each one ends with something you can hold.
No waterfall, and no six-month sprint. After every step you have a document or a running system, and you can stop there.
- 01
Understand
We meet your team, watch the process as it actually runs, and pull the data. You get a list of tasks AI can genuinely close, with numbers next to them.
- 02
Design
We take the one or two with the biggest payoff and work out how to build them: architecture, data, interface, and what we will measure.
- 03
Build
We build the pilot and connect it to your systems. We test on your real data, not on demo cases, and we fix what breaks.
- 04
Run it
We roll it out, train your team, and set up monitoring so you see problems before your users do. Then we hand it over, or keep running it for you.
The first stage starts within three business days of signature.
Three alternatives, and where we actually beat them.
We are not right for everyone.
- vs. consultants
We do not disappear after the deck
Consulting firms draw a strategy and leave. We go from the opportunity map straight into code, with the same team, all the way to production.
- vs. dev shops
We ask why before we build
A developer without business context builds what they were told. If something should not be built, we say so before you pay for it.
- vs. hiring in-house
Days, not six months
Hiring a senior AI engineer takes about half a year. We start in days, with a ready stack and dozens of similar problems behind us. When your own team is ready, we hand everything over.
Tell us what you want to change.
Thirty minutes on a call. Describe the process, and we will tell you whether AI belongs in it and roughly what it costs. If it is not our kind of work, we will point you at someone better.
A founding engineer answers, usually within one business day.
Got it. We will come back to you within one business day.
Three ways to start.
Scope is fixed at the front, so you never buy an open-ended retainer just to find out whether this works.
- 012–4 weeks
Audit
We map your processes, build an opportunity map, sort it by payoff and hand over a roadmap. You can take that roadmap anywhere, including away from us.
You want to know where AI will pay off, and where to start so the budget is not wasted.
- 028–12 weeks
Pilot
The audit, plus one solution taken all the way to production. Real users, measured results.
You know what to automate. You need a team that will actually build it.
- 036+ months
Partnership
Ongoing work with several streams running in parallel. A dedicated senior team that knows your business and keeps the context between projects.
AI is part of your plan for the next few years. You need a partner for that long.
What you get, in writing.
You own every line
All source code, fine-tuned weights and datasets are yours, in your repositories. No licence fee, no per-seat fee, no hostage code.
Your infrastructure, your data
We deploy inside your AWS, GCP or on-premise environment. Your data stays inside your perimeter. We have built SOC2-ready setups before.
No lock-in
You get deployment and API documentation, and training for your team. When you want to take the system in-house, you can.
Work starts in three business days
We are not going to spend the first month onboarding ourselves at your expense.
What people ask before signing.
What if it does not work?
During the audit we say plainly whether AI fits. If it does not, we point you at the process fix instead. During the pilot we agree the target metrics up front: if we miss them, we do not roll out, and you keep the code and the findings.
Why not just hire in-house?
Eventually you probably should. Hiring a senior AI engineer takes about six months, and we start in days. When your team is ready we hand over the code, the documentation and the training.
Will our data leak?
We work inside your AWS, GCP or on-premise environment. On one AI project the entire dataset stayed on the client's own server and only anonymised audio reached an external API. No names, no phone numbers.
Who owns the code and the model?
You do. Full source, deployment and API documentation, all IP rights, fine-tuned weights and datasets. That goes in the contract.
How much of our time does this take?
For the audit: four to six one-hour sessions with the people who actually do the work, plus access to the data. For the pilot: one technical contact and a 30-minute weekly check-in.
What if we do not have much data?
That is the normal starting point. We begin with what exists, plus public models and retrieval over your documents. Data builds up while the pilot runs, and fine-tuning comes later if it is needed at all.
Which languages do you work in?
Delivery, documentation and meetings in English, Russian or Kazakh. We have also built speech systems that handle Kazakh and mixed Russian-Kazakh speech, which most vendors cannot do.
Are you available, or is this a side project?
We deliberately take a small number of engagements at a time so that every project has senior people on it full-time. If your timeline does not fit our capacity, we say so on the first call rather than after the contract.