Projects Live · useproficio.com

Proficio

A personalized study-plan generator for Brazil’s ENAMED medical exam, built on a deterministic engine that ranks topics by their real historical incidence.

What it is

Proficio generates a personalized study plan for ENAMED — Brazil’s national medical exam, whose score also decides ENARE residency placement. A candidate sets how much time they have to study and rates their own confidence across seven subject areas; a deterministic scheduling engine turns that into a day-by-day plan, prioritized by how often each topic has actually appeared on past exams.

The ranking comes from real exam history: 1,925 questions across 18 past ENAMED exams (2011–2025), where the top five topics in each area account for roughly 62% of everything that’s been asked. The schedule runs in three layers — Core, Extension, and an Exam/Taper phase — with mock-exam sessions and a guessing-under-uncertainty protocol built in, and every plan exports as a PDF.

How it works

Calibrated to you

Your available study time and a seven-area self-assessment shape the plan before anything else.

Onboarding asks how much time you can dedicate and has you rate your own confidence, area by area, on a five-point scale — the schedule that follows is built around those two numbers, not a generic curriculum.

Weighted by real exam history

Topics are ranked by how often they’ve actually been tested, not by guesswork.

The ranking is built from 1,925 real questions across 18 past ENAMED exams (2011–2025). In most subject areas, the top five topics account for roughly 62% of everything that’s historically been asked.

Sequenced, not just listed

A three-layer schedule — Core, Extension, and Exam/Taper — replaces a flat topic list.

The engine spreads material across Core, Extension, and a final Exam/Taper phase, and schedules mock exams and a guessing-under-uncertainty protocol alongside the study material rather than as an afterthought.

The study plan

A day-by-day plan, up to exam day

A simplified view of the generated schedule: material is spread across Core, Extension, and a final Exam/Taper phase, with mock exams placed along the way toward the ENAMED exam.

Sample study plan

  • Day 1CoreClínica Médica
  • Day 9Mock examfull simulation
  • Day 22ExtensionPediatria
  • Day 38Exam prepguessing protocol
  • ENAMEDExam daytarget

study plan · demo

Features

Instant PDF plan

Every generated study plan exports as a downloadable PDF.

Mock-exam scheduling

Practice exams are placed into the schedule alongside regular study sessions.

Guessing protocol

An expected-value-based strategy for eliminating answers under uncertainty is built into the plan.

Progress tracking

The live site tracks completion against the generated schedule.

62%exam coverage from the top 5 topics per area
1,925real exam questions analyzed
18past ENAMED exams covered (2011–2025)
R$199starting price for a generated plan

Built in partnership.

Backend scheduling engine, exam-incidence dataset, and medical-accuracy sign-off — paired with front-end development and mentorship from Victor Barros.

Built in collaboration with

Victor Barros — Founder, investor, and mentor

Built Proficio’s front end and mentored the project’s engineering throughout.

Built with

  • TanStack Start
  • React
  • Tailwind CSS
  • Radix UI
  • Node.js
  • TypeScript
  • Python
  • Next.js
  • PostgreSQL (Neon)
  • Vercel Blob
  • Playwright
  • Anthropic Claude
  • Stripe
  • Vercel

Engineering

Deterministic scheduling core

The day-by-day plan comes from a fixed, rule-based engine, not a language model — the same inputs always produce the same schedule.

A Python forecast/value-model engine ranks each subject area’s topics by real historical incidence (1,925 questions across 18 past ENAMED exams, 2011–2025) and slots them into a three-layer sequence — Core, Extension, Exam/Taper — before any AI touches the output.

AI writes the words, not the plan

A language model only fills in the day-by-day written guidance on top of the fixed schedule — it can’t change what gets studied or when.

A schema-constrained layer (Anthropic Claude, validated against Zod schemas) generates the prose content for each day of the plan; the schedule itself is produced upstream by the deterministic engine, so a model hallucination can’t alter sequencing or topic weighting.

PDF, generated on demand

Each study plan renders into a real, downloadable PDF the moment it’s created.

Playwright drives headless Chromium server-side to render the generated plan into a PDF inside an on-demand Vercel Sandbox.

One monorepo, three runtimes

The frontend, the scheduling engine, and the API broker all live in one coordinated codebase instead of three disconnected ones.

A pnpm monorepo spans a TanStack Start/React frontend, a Next.js broker API, and a Python scheduling engine, sharing Zod-validated contracts; Postgres on Neon holds state and Vercel Blob stores generated PDFs.

Visit Proficiouseproficio.com

Have a dream project?

Tell me your idea and I’ll send you a quote.