Platform documentation

Overview, architecture, modules and frequently asked questions about the LAIFT platform and the work of Daniel P Francisco.

Overview

LAIFT unified platform: community governance (events, proposals with voting, tasks, team, messaging) and the "Learn" ecosystem (pharmacology and toxicology simulators, virtual clinic, laboratory and 3D anatomy atlas), with built-in artificial intelligence.

Explore the platform Request a demo

Architecture

1. 1

   Lightweight interface

   Fast web interface with no dependence on large frameworks, working on desktop and mobile. Learning modules share the same session, with no extra sign-in.

   • Web
   • Responsiva
   • Acessível
2. 2

   Edge API

   Service layer that only runs pre-approved actions and checks identity and permission on every call, close to the user for low latency.

   • API
   • Baixa latência
   • Permissões
3. 3

   Relational database

   Structured data with parameterized queries, versioned migrations, consent history and concurrency control for one-time actions such as voting, seats and check-in.

   • Dados
   • Migrations
   • Consistência
4. 4

   AI layer

   Language models serving teaching: virtual patient, preceptor, case generation and a platform assistant, with quotas, moderation and human review of what becomes shared content.

   • LLM
   • Cotas
   • Moderação

Design decisions

Each line starts from a real problem and shows the solution adopted.

Profiles and permissions

Who can do what, designed so that power and accountability go together.

• ProblemImproper access to management functions
  →
  SolutionThree base roles (visitor, member and administrator) combined with functional roles: manager, content reviewer, moderator, curator and mentor.
• ProblemRoles with too much power
  →
  SolutionCapability-based permissions, not title-based: each role unlocks only what the function needs, and several roles can be combined.
• ProblemCompromised management accounts
  →
  SolutionMandatory two-step verification for roles with power over people or content.
• ProblemIdentity forged in the browser
  →
  SolutionIdentity, role and account status verified on the server on every action, never trusted from the client.
• ProblemPower granted without a trace
  →
  SolutionRole assignment restricted to administration, with an audit trail for sensitive actions.

Security and privacy

Layered protection, from the browser to the data, with privacy as a design requirement.

• ProblemMalicious content typed by someone
  →
  SolutionAll text coming from people or AI is treated as plain text, and the browser receives a restrictive content policy.
• ProblemAutomated access attempts
  →
  SolutionAttempt limits on sign-in, sign-up and password recovery, with anti-bot protection where use without an account is allowed.
• ProblemPassword leaks
  →
  SolutionPasswords stored only as cryptographic hashes; resetting a password ends all open sessions.
• ProblemUse of data without consent
  →
  SolutionVersioned terms of use and privacy policy, with acceptance of each version recorded and preserved.
• ProblemExposed conversations
  →
  SolutionEnd-to-end encrypted messages between league members.
• ProblemExported spreadsheets with dangerous content
  →
  SolutionExports protected against formula injection.

Responsible AI

AI as a teaching aid, with limits on cost, content and privacy.

• ProblemCost and misuse
  →
  SolutionDaily quotas per person and per feature, with a global limit and automatic cut-off on failure.
• ProblemInappropriate content
  →
  SolutionModeration of questions and answers, with a human review queue for anything flagged.
• ProblemUnfounded answers
  →
  SolutionCurated content, cited sources and case generation with defined profiles and assessment criteria.
• ProblemMalicious instructions hidden in text
  →
  SolutionSeparation between user-written content and system instructions; nothing becomes shared content without review.
• ProblemExposure of personal data
  →
  SolutionQuestions logged in a minimized form without storing the raw text, and a clear notice about what is sent to the AI provider.

Learning, management and quality

From individual study to league management, with editorial and technical care.

• ProblemUnreliable anatomy content
  →
  SolutionPortuguese content sheets with human editorial review before publication in the atlas.
• ProblemStudy without follow-up
  →
  SolutionStudy statistics, study memory, a tutor and performance-axis assessment in the virtual clinic.
• ProblemAttendance fraud at events
  →
  SolutionSigned QR Code credential and a check-in terminal with permission-restricted actions.
• ProblemLeague decisions with no record
  →
  SolutionProposals with voting, events, collaborative tasks and an org chart inside the platform.
• ProblemRegressions and inaccessibility
  →
  SolutionAutomated tests, screen-reader accessibility checks and a staging environment separate from production.

Evolving

Directions the platform may take. The tag shows where each one stands; none is a promise with a deadline.

• In testing

  Tutor with study memory

  A tutor that remembers what a person has studied, only if they allow it, and adapts explanations and reviews. The memory is kept apart from other data and can be revoked or erased at any time, and never stores personal or health data.

• In testing

  AI personas with safeguards

  Assistants with defined roles (tutor, preceptor, advisor) and input and output rules: they refuse out-of-scope requests, never quote doses and leave a usage trail without storing the person's text.

• Under evaluation

  An assistant that reasons better over the library

  More precise selection of the passages that answer each question, intent detection (is it health-related? does it need a source? is it in the league's scope?) and continuous evaluation of answer quality.

• Under evaluation

  Assisted scientific search

  Search across open literature databases with a short summary and the source always in view. Anything stating practice, dose or interaction goes through human clinical review before it appears.

• Under evaluation

  Growing scientific library

  More articles in Portuguese, curation by area with public inclusion criteria, themed collections, favorites and citation in ABNT and Vancouver formats.

• In testing

  Atlas with a systems view

  Body navigation by system (cardiovascular, nervous, digestive and others), new content batches reviewed by the editorial board and integration with pharmacokinetics.

• Planned idea

  Applications and selection inside the platform

  Call for applications, sign-up, follow-up and results in one place, with openings and dates published from a single source.

• Under evaluation

  Sign-in with passkeys

  Simpler sign-in that is harder to phish, using the person's own device, for when the platform handles more sensitive data.

• Planned idea

  Partnerships and integration with other institutions

  Token-controlled access for those who want to consume content and aggregated data, and versions in other languages for leagues and students abroad.

A description of what was designed and implemented on the platform. It is not a certification or an external audit.

Modules

Frequently asked questions

Keywords

• Daniel P Francisco
• DPF Labs
• pharmacy education platform
• artificial intelligence health
• pharmacology simulator
• virtual clinic
• virtual laboratory
• 3D anatomy atlas
• toxicology
• HealthTech
• LAIFT
• digital identity services
• scientific writing
• health web development