Jakimio.
Applied AI for real products

AI that acts —
and proves every step.

We build large language models into the products themselves: agents that automate real business processes, an ambient scribe that turns a conversation into a structured document, and apps you can just talk to. Never a bolt-on chatbot — woven in, with a human in the loop and a full audit trail on every decision.

◆ Model-agnostic ◆ Tool use via MCP ◆ Human-in-the-loop ◆ Cost-traced per run
See the work → Start a conversation
The common foundation

Four principles under everything we ship

The same spine runs through every product. It is what makes an AI feature safe to put in front of a clinician, an auditor, or a regulator.

01 / PROVIDER

Model-agnostic

Every product works with any LLM provider. The model is chosen and configured per deployment — never locked to one vendor.

02 / TOOLS

Tool use via MCP

The AI reaches real data and actions through the Model Context Protocol — it reads records and does the work, not just talk.

03 / OVERSIGHT

Human in the loop

Approval gates on anything that writes or acts on sensitive data. Low-confidence results route to a person, not into the record.

04 / EVIDENCE

Traced & costed

Every run traced end to end, with token cost captured per run, per model, per workflow. Nothing the AI decides is a black box.

Selected work

Three products, one discipline

Real products in regulated, audit-heavy work — each putting AI inside the app itself, with a human in the loop and every decision traceable.

AI platform Business automation

Automation Studio

A visual canvas for building agentic workflows that automate real business processes. Drag deterministic steps and AI-agent nodes onto one board and wire them together — non-developers compose real automations in minutes, not engineering tickets.

How the AI works: the diagram compiles into an executable agent graph. Each agent runs a reason-and-act loop with its own model, tools, memory and knowledge base; short-term memory is checkpointed so a run can pause for approval and resume. Writes are gated, and every decision is logged with the data it touched, its reasoning, and its exact token cost.

StackVue Flow canvas · agent runtime · MCP servers
Per agentModel · tools · memory · knowledge base
EngagementProduct design & build
Status● In active development
Speech AI Clinical & non-clinical

Business Speech-to-Text

A speech-to-text and structuring platform for any domain — clinical consultations, business meetings, field reports, legal. It listens, transcribes, and turns the raw conversation into a structured document that follows your template.

How the AI works: two stages. A provider-agnostic speech-to-text engine turns audio into a verbatim, speaker-labelled transcript; then a language model turns that into a structured document — sectioning to a chosen template (SOAP for a clinical note, minutes-and-actions for a meeting), handling negation, filtering small talk, and synthesising details mentioned minutes apart. A person reviews and signs; the exact template, model and version are frozen onto every document for traceability.

StackSTT (Deepgram / Azure / Google / Whisper) · any LLM
DomainsClinical · meetings · field service · legal
OutputStructured document + summary (FHIR for clinical)
ModesStandalone · embedded · API / SDK
Status● In production
HR app + AI assistant

HR App

An HR app people actually enjoy using — because you can just talk to it. Alongside the dashboard, leave and absence, approvals, a team calendar and employee records, there’s an AI chat built right in: ask “how many days do I have left?”, “request three days off next month”, or “who’s off this week?” — and it just does it.

How the AI chat works: it connects through its own MCP server, so it reads live data and performs real actions inside the app’s own permissions — booking leave, checking balances, even charting headcount on request — not a chatbot bolted on the side. The provider and model are set by an administrator at runtime; a deterministic mock stands in during automated tests.

The chatBooks leave · checks balances · charts trends — in plain language
AlsoDashboard · leave & absence · approvals · team calendar · employees · equipment
RolesEmployee · manager · administrator
StackKeycloak auth · dedicated MCP server · any LLM
Status● Shipping
hr-app · assistant
The HR app AI assistant panel, in dark mode, answering in plain language and rendering a bar chart of approved leave days per month
The AI chat — ask in plain language; it reads live data and even charts it.
hr-app · dashboard
HR app manager dashboard in dark mode: leave balance, pending approvals, who is out this week, active employees and equipment
Manager dashboard — balances, approvals, headcount and equipment.
How we deliver

Built to hand over, not to hold hostage

The same terms we put in writing for our enterprise clients. They are why a risk-averse buyer can commission a custom platform without betting the business on a vendor.

01

Fixed-price, milestone-based

Work is delivered in phases, each signed off in writing before the next begins. Payment follows the sign-off — not the promise.

02

100% of the IP is yours

Code, schemas and documentation transfer to the client on final payment. No proprietary framework, no dependency you cannot replace.

03

Open-source only

Every dependency is MIT- or Apache-2.0-licensed, with a software bill of materials delivered as project documentation.

04

Auditable by construction

Append-only audit trails, version-controlled migrations, and approval-gated deployments — the change-control evidence regulated work requires.

The studio

Jakimio Technologies

We are a software studio building end-to-end software solutions and AI integrations — we design, build, test and deploy the whole system, from the data model to the last pixel. We work where correctness matters: regulated, audit-heavy domains where a wrong answer has a cost.

Our edge is putting AI to work responsibly inside real products — with the governance, observability and human oversight those domains demand. The products here are the proof.

WHAT YOU GET
✓  End-to-end delivery — design, frontend, backend, deploy
✓  AI woven into the product, not stapled on
✓  Governance & audit as a first-class requirement
✓  Clean handover — your code, your infrastructure
HJ
Hristijan Jakimovski
Founder · Jakimio Technologies
Contact

Tell us what has to be right.

A new build, a system that needs governance put around it, or a partnership — start here. Investment and partnership enquiries welcome.

+389 70 254 801Hristijan Jakimovski
info@jakimio.com
29 Noemvri 57, 1000 SkopjeNorth Macedonia