Product

Every module, and how it actually works

Oskela is nine training tools sharing one taxonomy, one drug database and one analytics layer. Here is what each one does.

Start free See pricing

Clinical case simulator

A case opens with a presentation and a set of observations. Every decision you make: a question asked, a test ordered, a drug given: applies a bounded change to the patient's physiology and moves the encounter to the next stage.

Live vitals. Pulse, blood pressure, saturations, respiratory rate and temperature respond to what you do, with colour-coded thresholds.
Branching. Stages can jump based on the option chosen, so two students can take genuinely different routes through the same case.
Critical failures. Certain choices: a beta-blocker in acute severe asthma, for example: are marked critical and fail the attempt regardless of the rest of the score.
Domain scoring. History, examination, investigation, diagnosis, management and safety are scored separately against a stable denominator, so feedback points at the weakness.

Prescription writing lab

You write a real prescription: drug, dose, route, frequency, duration: for a patient with a chart, a weight, a renal function and an allergy list. It is then checked by a deterministic rule engine.

The safety engine is not an AI. Dose ranges, renal thresholds, allergy cross-reactivity and interaction severity come from the drug database and fixed rules. A language model may explain a finding afterwards, but it never decides whether a prescription is safe.

Expected-drug groups accept clinically equivalent alternatives rather than one "right" answer.
Renal adjustment is checked against the patient's calculated eGFR band.
Interactions are surfaced against both the new prescription and the patient's current medicines.
Rules scoped to a drug are skipped when that drug was not prescribed, so the denominator stays honest.

Assessment you can defend

Everything that determines a mark happens on the server, where the browser cannot reach it.

Server-authoritative timing

The deadline is computed and stored when the attempt starts. Closing the tab, changing the clock or replaying a request cannot buy time.

Per-attempt shuffling

Question and option order are shuffled per attempt and stored, so the sequence is reproducible for review but not shared between candidates.

Answers stay hidden

Correctness is never sent to the browser while an attempt is in progress: it appears only after submission, subject to the result visibility rule.

Item analysis

Facility, discrimination and option-level distractor analysis across the cohort, so weak items can be retired.

Where AI is used, and where it is not

AI is used for teaching, drafting and explanation. It is never the authority on safety or on a mark that counts.

Tutors answer questions, compare drug classes and generate practice items on demand.
The viva examiner conducts an adaptive oral examination and scores it across defined domains.
Content generation drafts questions and cases: always into a review queue, marked as AI-generated, never straight into the published bank.

Deterministic engines own prescription safety, question scoring, exam timing and checklist marking.

See it with your own curriculum

Bring a syllabus and a cohort: we will show you the modules mapped onto it.

Get in touch Start free instead