Services Healthcare Technologies Blog About us Contact Write to us

We build any software. Our subspecialty is medicine.

Consdinamic designs and builds applications, AI assistants and data platforms for companies and institutions. In healthcare we go deepest: our own products, used daily in clinics, and custom solutions for hospitals and public institutions.

Since 2008, Republic of MoldovaIn the EU through Softmed Labs SRL, RomaniaProducts in Romanian, Russian and English
We already work with companies in the medical field
Gral Medicalprivate medical network, Romania · 29 locations
AiforiaAI for pathology, Finland
Private clinics in Romaniawebsites, maintenance, search visibility

From an idea on a sheet of paper to a product used every day

We work for any field: platforms, AI assistants, automations, data and websites. Every project starts from your processes, not from our catalogue.

Web and mobile applications

Platforms, portals and applications for customers, patients or internal teams, from the first sketch to day-to-day operation.

AI assistants: voice and chat

Assistants that answer the phone and chat, in Romanian, Russian and English, connected to the systems you already have.

Automations and integrations

We connect your existing systems: appointments, payments, telephony, reports. Less manual work, fewer errors.

Data, reports and AI

We prepare the data, build reports that are easy to read and AI solutions that work on your data, in line with data protection rules.

Websites and online presence

Fast websites, easy to find in search, in several languages.

Digital products, from idea to launch

In 2026 we launched over 30 digital products, each with its own website, online payments and measurement. We bring the same pace to custom projects.

Illustrations with sample data
De-identified dataresearch agreements Training and validationseparate sets Regulated frameworkclinician · accredited laboratory The hospital remains the controller of its data

We build neural networks for real problems in healthcare

We train models on de-identified data, provided by hospitals under research agreements: digital pathology images, clinical information, signals from conversations with patients. The model learns to recognise patterns; the decision remains the doctor's.

Every model follows the same path: data prepared and de-identified, training and validation on separate sets, then use only within the regulated framework, together with accredited laboratories and clinics.

What a training run looks like: loss falls, accuracy rises

The curves below are generated in the page, with sample data: at each epoch the model sees the training set again, the weights are adjusted, and the loss on the validation set tells us when to stop.

Training lossValidation lossAccuracy
Illustration with sample data, drawn in the browser

In medicine, writing good code is not enough

You need to know how a clinic, a call centre and a laboratory work, what is permitted with a patient's data, and where the software stops and the doctor's decision begins. This is where we go deepest.

Review management platform

Reads the Google reviews of every location in a medical network, drafts a personal reply in the patient's language and publishes it once a member of staff approves it. Each month, every location receives a report: what patients appreciate and what needs improving.

Gral Medical, a private medical network in Romania: 29 locations, over 15 months in use.The share of reviews that receive a reply rose from about 30% to 100%.

AI voice assistant for appointments

Answers the phone, understands what the patient needs, checks the doctor's free slots in real time and books the appointment. When the case calls for a person, it transfers the call to an operator, together with the context.

FAQ assistant (chatbot)

Answers patients' frequent questions, day and night, from the institution's approved information. It gives no medical advice and hands the conversation over to staff when a question goes beyond its role.

In-clinic wayfinding by QR code

The patient scans a QR code at the entrance and sees the route to the right consulting room on the floor plan, with step-by-step directions. No app to install and no new equipment.

DNA screening system

An engine that reads genetic data and prepares a structured report for the clinician on how the patient's genes influence their response to medicines. Designed first for oncology.

Oncology data for AI training

A collaboration with Aiforia (Finland), a company developing AI for pathology: datasets of digital pathology images and clinical information. The data comes from hospitals under research agreements, is de-identified, and the hospital remains the controller of its data.

The patient app of a medical network

We are part of the working group that designs and develops the Gral Medical patient app: appointments, results and communication with the clinic.

Clinic websites and online visibility

Websites for private clinics, with maintenance and ongoing work on search visibility, so that patients find the right service and the right doctor.

Illustrations with sample data · Solutions intended for clinical decisions are used only within the regulated framework, together with accredited laboratories and clinics.

Code you can read, decisions you can trace

Every function of an assistant leaves a trace: what it understood, what it checked, when it handed the conversation to a person. Below, a change that is real in form, with sample data.

Illustration with sample data. No real patient, no real system.

What we can build together

Project ideas with ministries, hospitals, insurers and industry partners. The list is open: we build to order, starting from the partner's needs.

01

Bilingual AI phone line for appointments

A voice assistant in Romanian and Russian for hospitals and polyclinics. It takes calls at any hour and shortens the queue at the front desk.

02

Assistant for citizens using new digital services

When the electronic health record or the electronic prescription is launched, citizens and doctors have questions. An assistant answers them from official information.

03

Patient feedback for public hospitals

A single system that gathers what patients say about each hospital and turns it into a monthly report for management.

04

Wayfinding in large hospitals

QR codes at the entrances and on every floor guide patients to the right department. In one building, the first version is ready within a few weeks.

05

Data prepared for secondary use

Digitising pathology archives and preparing de-identified datasets for research and AI, in the direction of the European Health Data Space.

06

Genetic screening in oncology

Together with an accredited laboratory: genetic information that supports the oncologist in choosing a safe dose for each patient.

Well-known tools, chosen for each project

We do not have one stack for everything. We choose what fits the systems you already have and the people who will maintain the product after us.

Languages and frameworks
  • Python
  • TypeScript / Node.js
  • React
  • Next.js
  • Tailwind CSS
  • Static HTML / CSS
Artificial intelligence
  • PyTorch
  • OpenAI API
  • Anthropic Claude API
  • Google Gemini API
  • Voice models (speech-to-text, text-to-speech)
  • RAG on your own documents
Data and infrastructure
  • PostgreSQL
  • SQLite
  • Docker
  • GitHub Actions
  • Matomo (cookie-free analytics)
  • Playwright (automated tests)
Integration and operations
  • n8n (automations)
  • SIP telephony / call centre
  • Stripe (payments)
  • Google APIs (Search, Analytics)
  • Webhooks and REST APIs
  • Hosting in the EU

First we listen, then we show something real

01

First we listen

We start from your problem, not from our catalogue: the processes and constraints you have.

02

A working prototype

Within two to three weeks we show something real that people can try, so the conversation is not about slides.

03

Start small, then extend

We begin with a first version used by real people, with indicators agreed in advance. Only what proves its value is extended.

The institution stays in control

01

The institution stays in control

The clinic or hospital remains the controller of its data. We work as a processor, under a data processing agreement, in accordance with the GDPR.

02

As little data as possible

Each product uses only the data it needs. For research and AI we work with de-identified data, under research agreements.

03

A person approves

Text drafted by AI on behalf of an institution is read and approved by its staff before it reaches patients.

04

Support, not diagnosis

Our assistants help with communication and organisation. They do not make diagnoses. Tools intended for clinical decisions follow the regulatory pathway required for them.

Since 2008, technology anyone can use on their own, first time

Consdinamic was registered on 1 April 2008 in the Republic of Moldova. We started with self-service payment terminals for the public: technology that anyone must be able to use on their own, first time. Today we build custom software and AI, with a subspecialty in the medical field: our own products, used daily in a private medical network, and solutions built around each partner's processes.

In the European Union we work through our sister company in Romania, Softmed Labs SRL, so partners can contract us in Moldova or in the EU.

CompanyConsdinamic SRL
CountryRepublic of Moldova
Registered1 April 2008
In the EUSoftmed Labs SRL, Romania
FounderVladimir Burlac
LanguagesRomanian, Russian, English
Products launched in 2026over 30
Locations running the review platform29

What we have learnt building for healthcare

All articles

Tell us what you need. We come back with a prototype.

Write to us in Romanian, Russian or English.

Vladimir BurlacFounder contact@soft-med.ro +40 749 620 127 Română · Русский · English

Write to us