Dina Kalibekova

Healthcare · AI · Two countries

Rette

Description

I joined Rette as an intern when it was a simple MVP with no customers. Today I lead Defect AI, our product for clinics in Kazakhstan, used by 30 clinics and a medical university, and work on Rette, its sister product for the US. My work, in my own words, is to close all the holes.

Project details

Date:
January 2025 → now
Team:
3 founders, about 5 developers, my interns
My role:
Founding team member
01

Overview

Where I started, and how the work grew.

In numbers

300+
conversations with doctors, calls and clinic visits
3
first clinics, sold by me while it was still an MVP
1 → 10
modules in Defect AI, plus solutions for single clinics
30 + 1
clinics and a medical university using Defect AI today
$20,000
grant from Social Startup 2025, a pitch I led
~1 week
from a first conversation to a working model

The journey

  1. January 2025Defect AI · Kazakhstan

    Intern, at the MVP

    The fourth person in a company with a simple MVP and no customers.

    MoreLess

    Three founders, a very simple MVP, no customers, and one medical expert helping us build it. I was the fourth person in the company, in my third year of maths at Nazarbayev University.

  2. 2025Defect AI · Kazakhstan

    Talking to doctors

    Called and visited doctors, and sold our first three clinics while it was still an MVP.

    MoreLess

    We ran targeted ads on Instagram and leads started coming in. I called the doctors, asked where the program failed them, and brought the gaps back to the team. I went to clinics in person, and sold our first three customers while it was still an MVP.

  3. 2025The company

    Covering every gap

    Analytics, design, support, sales and applications, while still studying.

    MoreLess

    Three of us in the office, a founder in the US, a developer in Berlin. I helped with analytics, design, support, sales and accelerator applications, while still studying.

  4. 2025Defect AI · Kazakhstan

    A grant I won

    Took Social Startup 2025 from the first application to the final pitch: a $20,000 grant.

    MoreLess

    On my own I applied to Social Startup 2025, run by the NU Impact Foundation, and took it from the first application to the final pitch, about how much of a doctor's day goes on checking records by hand. Defect AI won a $20,000 grant.

    Dina pitching on stage in front of a slide about doctors' time spent on manual checks.
    The final pitch at Social Startup 2025.
  5. End of 2025The US market

    Two months in the US

    TechCrunch Disrupt, Silicon Valley Residency, and StartX's medical track.

    MoreLess

    We exhibited at TechCrunch Disrupt in San Francisco, and I made the merch and presentations for the booth. We went through Go Abroad, Silicon Valley Residency with Alchemist's support, and StartX, Stanford's accelerator, which then took us into its medical track.

  6. 2025 – 2026Defect AI · Kazakhstan

    From one module to ten

    I interviewed doctors on how they check their records, and one module became ten.

    MoreLess

    Our first model checked inpatient records only, a small part of a doctor's paperwork. I interviewed doctors about how they check their records, against which protocols and in what format, and one module became ten, plus solutions made for single clinics.

    The Defect AI case study →
  7. 2026Both products

    Leading one, building the other

    Took over Defect AI, and researched US healthcare for Rette, now in the Mayo Clinic accelerator.

    MoreLess

    The founders moved to the US product, so I took over Defect AI: the features, the developers, the doctors' needs and the design. Defect AI was named Startup of the Year at the AI & Digital Bridge Awards. For Rette I researched how US healthcare works, and we were accepted into the Mayo Clinic accelerator.

  8. NowBoth products

    Closing the holes

    Automated the routine checks, and spend that time on Rette's partnerships and sales.

    MoreLess

    Analytics, process, films, exhibitions, and my own interns. I automated the routine checks, so errors and unusual usage flag themselves, and spend that time on Rette: partnerships, and how to sell it.

Dina with two teammates at the booth at TechCrunch Disrupt.
With the team at TechCrunch Disrupt, San Francisco.
Dina on stage holding a Social Startup 2025 Winner sign for Defect AI.
Social Startup 2025: Defect AI among the winners.
Dina working at her laptop, chin on her hand.
Getting ready for Silicon Valley Residency's demo day.
02

Defect AIKazakhstan

Our product for clinics in Kazakhstan, which I lead: how I design it together with doctors.

Designing from interviews

The Defect AI landing page and the platform itself came out of user interviews: I designed both from what doctors told me.

The Defect AI landing page: a headline about cutting defects in discharge summaries by 90%, with photos of a doctor at a laptop and a checked document.
The Defect AI landing page, defect.kz (Kazakhstan).

How I build a new module

How I think when I make a new module for Defect AI, from a doctor's problem to every clinic.

Not helpful yet? Back to the sketch.

  1. 01

    Listen

    I talk to clinics, or look for the problem myself.

  2. 02

    Sketch

    Once there's an idea, I sketch it on paper and a board.

  3. 03

    MVP

    A rough version, fast.

  4. 04

    Test

    Doctors try it. Does it really take the pain away?

  5. 05

    Make it simple

    Time just for ease: many doctors are older and not at home with tech.

  6. 06

    Final product

    Simple, useful, and only then shown to everyone.

  • The office whiteboard in July: a numbered list for Defect AI, from a landing page and analytics search to interviews with clinics.

    01July on the office board: interviews with clinics, and what to build next

  • The whiteboard in August: a platform redesign, templates in each doctor's own style, and what stats to give doctors.

    01August: templates in each doctor's own style, and what stats doctors need

  • The whiteboard in November: a new module, improving templates, and testing which AI models to move each module to.

    01November: a new module, and testing which AI models to move to

  • A pen sketch of the account flow: a user opens the site, gets a personal account, and either uses it alone or creates a clinic.

    02The account flow, on paper: one doctor, or a whole clinic

  • Pen sketches of the side menu for each kind of account: active, trial and expired.

    02The menu for every kind of account, before any code

  • A screenshot of the analytics page with my red notes: change the order of the periods, swap two charts, fix a scale.

    05My notes on the analytics page: what to reorder, swap and fix

1 → 10 modules, plus solutions made for single clinics.

a module, in every clinic made for one clinic

03

RetteUnited States

Its sister product for the US, where I first had to learn a new healthcare system.

One problem, two countries

None of us lived in the US or knew its healthcare system, so before we could build for it I had to learn how it works, and how it differs from Kazakhstan's.

Kazakhstan

  • One central public system with shared national rules, plus local law.
  • Most people rely on public cover; private insurers are few.
  • Fewer layers, so the rules a record must follow are clearer.

United States

  • Public programmes like Medicare and Medicaid, and many private insurers.
  • Each insurer has its own policies, and they can stack: one part of a procedure is covered by one, the next by another.
  • Checking records sits inside revenue cycle management.

The problem is the same in both: check the documents and catch the errors early. In the US it's a thousand times more complex.

A pen sketch of a US claim: patient, clinic, physician, medical coder, billing, insurance, with a note that Rette sits at the coder.
My sketch of how a US claim travels, and where Rette fits (Rette · United States).
04

Everything I've made

Most of my work is small and easy to miss, and it isn't only product. Pick a kind of work to light up every piece of it, or hover any piece to see where it sits.

05

When I have time, I work visually

Films, merch and banners: the visual side of the work, made between everything else.

The clinic film, 2:32, for clinic directorsDefect AI · Kazakhstan
The booth T-shirt design: on the front, defect.app and 'reduce claim denials with healthcare AI audit'; on the back, a QR code and hashtags.
The booth T-shirt, front and back, 2025
Six T-shirt designs for Rette: a REVIEWED stamp, 'Every single line', 'The AI that actually reads the chart', a pulse turning into a check, a code diff of missed and caught, and a spine with one vertebra circled.
Rette merch: 6 of 27 back-print ideas, 2026Rette · United States
Asel, the film's doctor character, drawn with different expressions.
Asel's expressions, drawn for the filmDefect AI · Kazakhstan
A frame from the Rette paper plane film.
The paper plane film, in 3DRette · United States
06

Reflection & next steps

Where it stands

Defect AI is used by 30 clinics and a medical university in Kazakhstan. Rette is in the Mayo Clinic accelerator in the US.

What I learned

One thing this job taught you, about doctors, about building, or about yourself.

What's next

What you're building next for Rette, and how you'll know it works.