A software and training company from Tăuții-Măgherăuș, Maramureș, Romania, since 2006

Digital laboratories on synthetic data, for teams that work with artificial intelligence.

Virtual Soft builds training applications and laboratories on the real processes of a sector — with synthetic data, provoked failures and verifiable results. One of them runs below, directly in your browser.

Synthetic data generator Functional — runs in your browser

Generating…

  • generated
  • target distribution

Check results

  • 4 working demos
  • Runs in your browser, no server
  • Synthetic, reproducible data
  • Source code delivered

What Virtual Soft does

Virtual Soft S.R.L. is a software and training company based in Tăuții-Măgherăuș, Maramureș County, Romania, founded in 2006. It builds digital laboratories on synthetic data, web applications and applied AI training programmes, and provides IT consulting for AI adoption. Each service line starts from the client's real process and ends with something that can be run, checked and taken over by the client's team.

Software applications for training and digital laboratories

Laboratories built on a sector's processes, where learners work with synthetic data and provoked failure cases.

  • Synthetic, reproducible datasets
  • Provoked failures and portfolio results
  • Source code and a perpetual, transferable licence

Applied AI training programmes

Online programmes for groups, organised in theory hours and laboratory hours.

  • Course materials and laboratory sheets
  • Assessment instruments
  • Exercises on the team's own processes

IT consulting for AI adoption

Process analysis starting from its constraint, judged on cost, confidentiality and control.

  • Inference architecture, local or EU API
  • AI usage policies
  • Measurable pilots

Full description of the services

Demos

Every demo runs in your browser, on synthetic data or on illustrative values labelled as such, and shows its method. The images are computed from the same models the demo uses.

  • Functional — runs in your browser

    Synthetic data generator with validation

    Generates, in your browser, a synthetic dataset for a real business process (SME property insurance underwriting or SME credit dossier review) and checks it immediately with seven controls computed on the generated data: schema, referential integrity, temporal coherence, target distributions, correlations, injected anomalies and the absence of personal data.

  • Functional — runs in your browser

    Laboratory with a provoked failure

    Shows how a laboratory exercise is built: the learner receives a synthetic underwriting document, runs a simulated extraction pipeline (read, extract fields, check rules, decision), finds the stage where an injected failure stops it, chooses a diagnosis, decides whether the discrepancy must be escalated or tolerated, applies the fix and exports the log as a portfolio item.

  • Functional — runs in your browser

    The process constraint

    Shows, on an editable claims process, that the output of a workflow is set by the stage with the lowest capacity. The default calculation is deterministic: it finds the constraint, the system throughput and the queue that builds up in front of it, then compares automating an arbitrary stage with automating the constraint. An optional mode adds variability in arrivals and processing times and shows the average queue that forms in front of nearly full stages too.

  • Functional — runs in your browser

    Cost per task for model configurations

    Compares the cost per task and the monthly cost of up to three configurations, through an API or with local inference, starting from the monthly volume and the tokens per task. It shows which configuration costs least at the entered volume, computes the break-even between a local and an API configuration, first against the cheapest one, and shows how the cost changes with volume.

Method

Five stages, in this order, for every project.

  1. Analysis

    The client's real process, its constraint and the result that has to be measured.

  2. Design

    Laboratory scenarios, synthetic data and assessment criteria, validated with the client.

  3. Development and pilot

    Shared engine plus domain pack, tested with a pilot group.

  4. Delivery and transfer

    Source code, documentation and knowledge transfer to the client's team.

  5. Support

    Fixes, content updates and assistance for the team that takes over the system.

The approach in detail

Projects developed

Applications developed by Virtual Soft, through Olos Liviu Virgil.

2024–2025

QuickGuee

Dossier management for the one-stop energy-efficiency office of a Romanian county council — email and Excel import, duplicate and status checks, reports.

Web applicationAutomation
2025

Property titles

Property title requests for the Maramureș County Prefecture — email intake, two-reviewer evaluation, signed reports, submission to OCPI.

Public administrationApproval workflow

All projects

Company details

The company can be checked publicly with ONRC and ANAF.

Name
Virtual Soft S.R.L.
CUI
18736264
Trade Register no.
J2006000901248
Registered office
Tăuții-Măgherăuș, Maramureș
Main activity
8559 — Other education n.e.c.
Founded
2006

Olos Liviu Virgil — Owner. Engineer in Automation and Computers, with more than 25 years in IT and AI consulting.

All identification details

Shall we talk about a laboratory?

For an offer, a demonstration or a technical question, write or call. Office hours 8:30–16:30 (Romanian time).

How we work — what to send, what the offer contains