Enterprise Medical Platform for Asklepiy
2,500+ employees. 1M+ patients annually. What started as a basic CRM became the operational backbone of Asklepiy — one of the country's largest private medical networks, spanning four clinical brands.
We grew the system as the business grew — module by module, integration by integration. Today the platform unifies what used to be seven separate tools into one interface, serving Asklepiy Medical Center, Asklepiy Family, Asklepiy Recovery, and Asklepiy Beauty & Health on a shared core.
One core, four brands
One platform core. Four clinical brands.
Asklepiy Medical Center, Asklepiy Family, Asklepiy Recovery, and Asklepiy Beauty & Health run on a shared platform core — one data model, one role system, one integration layer. External systems connect through APIs; brands differ at the surface, not in the foundation.
Platform evolution
Ten years, module by module.
The platform was never rebuilt — it accreted. Each module shipped into a living system, on the same core, without interrupting daily operations.
What we built
01
Adaptive role architecture
Architecture for 10+ user roles with adaptive workspaces. One employee can hold multiple roles at once — the system reshapes to context instead of forcing account switching.
A duty physician who also heads a department doesn't log out and back in. The workspace reassembles around the active role — navigation, data scope, permissions — in one click.
One person, many roles. No fragmentation.
| Direction | Visits | Plan | % | Status |
|---|---|---|---|---|
| Medical Center | 4 182 | 4 500 | 93% | On track |
| Family | 2 957 | 3 100 | 95% | On track |
| Recovery | 1 248 | 1 600 | 78% | Below plan |
| Beauty & Health | 684 | 1 050 | 65% | Needs attention |
02
Third-party integration
Third-party medical systems integrated via API: lab data, research, financial records, operational data — surfaced in one place.
Lab results, research data, financial records and operational feeds arrive through APIs from systems the network already runs. The platform doesn't replace them — it makes them legible in one place.
Seven tools, one surface.
| Date | Status | Duration | Comment |
|---|---|---|---|
| 13.05 | Completed | 35 min | Scheduled consultation, follow-up booked |
| 06.05 | Completed | 20 min | ECG reading, no remarks |
| 29.04 | Rescheduled | — | At patient's request, new date 06.05 |
| 22.04 | Completed | 45 min | Initial appointment, examination prescribed |
| 15.04 | Cancelled | — | No-show, reminder sent |
Systems integration
A two-way line into medical infrastructure.
The platform runs a bidirectional integration with the network's medical information system — in production since 2021, across an API surface of 78 endpoints. Clinical and operational data move both ways: the system reads what the network records and writes back what it decides, without either side becoming the other's bottleneck.
03
Field operations
Medical representatives work across regions — visit planning, route control, and plan-versus-actual tracking per facility. Regional management sees execution in real time.
A representative sees their day before it starts; the network sees every visit after it happens.
Distributed teams, single source of truth.
04
Real-time messaging, built in.
A full internal messenger lives inside the platform — direct, group, and task-bound conversations, reactions, quoting, media and voice messages, read receipts. Drafts and an offline send-queue mean nothing is lost when the connection drops.
Any message turns into a task in one step; every task carries its own thread. The conversation and the work it produces stay in the same place — no second tool, no copy-paste between chat and tracker.
Talk and work, one surface.
05
AI as a digital employee
The AI layer operates in three modes. Analytical: the user asks, the system answers — dashboards, a catalogue of ready analytical presets, visit briefs, cross-source questions answered in seconds. Agentic: the user delegates — tasks, reminders, multi-step sequences that run for weeks. Proactive: the system initiates — morning analytics before the team arrives, flagged deviations, rescheduling proposals.
It was designed read-only first: the model could show and suggest, but every action required a human. Once the system proved itself, the client extended its mandate — an autonomy matrix now defines which actions run autonomously and which require confirmation, with a full audit trail on every action and one-click scenario stop. Deleting data or writing to external systems is impossible at the architecture level.
Privacy is structural, not promised: zero patient names ever reach the model. A centralised anonymisation service processes every data response before it returns to the AI — with full test coverage. Field representatives see only their own portfolio, enforced at the data layer.
The deterministic foundation does the counting; the model interprets. Dashboards, aggregates, and triggers are computed by code — the AI formulates, deepens, and acts on them. The layer is entering production through 2026.
Weeks of manual reporting, removed.
06
Role-aware mobile
Mobile application with role-aware functionality — each role sees its own interface, not a stripped-down version of the desktop system. Shipped as SalesCure, the platform's public product layer.
Native to each role, not a desktop afterthought.
Platform scope
What ten years of modules adds up to.
Engagement model
Two principals. Ten years. One system.
Dmytro Skoropad leads product strategy and design. Igor Ratsun leads architecture and engineering. Senior specialists join per module — the principals stay. Ten years of platform decisions made by the same two people who made the first one.
— —
Continuously deployed since 2016. Ten years of engagement — not one of them a rewrite. The longest engagement in our portfolio.
Considering a complex platform engagement?
Start a discovery call