PTEN
Back to portfolio
Open Application
Proprietary product, prompt engineering

Lapidador

An AI tutor that guides users from a loosely defined need to a ready-to-use prompt and explains the choices made along the way.

Problem, decision, and evidence

Problem

People who understood prompt theory still struggled to turn a real request into a clear instruction.

Role and decision

I designed the product, conversational experience, and technical architecture. The central decision was to gather context before applying techniques from the catalog.

Available evidence

The application is live and the full flow can be inspected, from the initial request to the annotated output. No public benchmark for performance or adoption is available.

Product: the entry point where users describe the need they want to turn into a prompt.
Product: the conversational step that gathers context before recommending a structure.
Product: the organized prompt delivered after the user answers the contextual questions.
Product: the instructional explanation of the techniques and decisions applied to the final version.

Click to enlarge

Scope of work

Role

Product, experience architecture, and technical direction

Audience

Professionals who need to improve prompts at work

Methods & technology

React, TypeScript, Express, AWS, OpenAI, and Groq

Outcome

Users move from a rough idea to a structured prompt with an explanation of the choices

Context

A proprietary product created around a recurring classroom difficulty: knowing techniques was not enough to turn a workplace need into a useful instruction.

Business problem

The number of available techniques could make prompting feel more complicated. Learners needed guided practice that helped them formulate the request, delivered a usable prompt, and showed why it had been structured that way.

Audience & personas

  • Business professionals who use AI at work
  • Students in AI courses and workshops
  • People who need better prompts without studying the technical taxonomy

Adoption barriers

  • Difficulty explaining context to AI
  • Prompts that are too generic
  • Dependence on templates without knowing how to adapt them

Solution design

  • Conversational flow that asks for context when information is missing
  • Catalog of 23 techniques applied according to the stated need
  • Final response with a ready prompt and instructional explanation
  • Authentication, credits, and payments to support product access

Scope and deliverables

  • Catalog of 23 prompt engineering techniques
  • Prompt structure recommendation engine
  • Landing page, authenticated application, and payment flow
  • First publishable version built in two weeks

Evidence and signals

  • Live application available for direct inspection of the flow
  • Demonstration of contextual questions, output, and technique explanations
  • No public usage, retention, or performance improvement data

Lessons learned

Explaining the choices helps users adapt the prompt later instead of depending only on the ready-made answer.

Next improvements

  • Add examples by role, version history, and a corporate mode with a team prompt library.
  • Build an MCP server to integrate Lapidador's engine with chatbot tools.
View implementation details
  • The first publishable version, including the landing page, AI engine, and payment flow, was built in two weeks. That timeframe describes the first version, not a finished product.
  • The implementation includes a Cognito-authenticated application, credits and Mercado Pago payments, DynamoDB, Express, ECS Fargate, and CloudFront distribution.
  • The catalog contains 23 techniques. This number describes the integrated scope, not a business impact or learning outcome.

Related services

AI Prototypes & DemosMVPs, automations, and demos to test an idea's value before scaling — using tools like ChatGPT, OpenAI API, n8n, TypeScript, and AWS to prove what works with a tight scope.Enablement Playbooks & AssetsEnablement means giving a team practical conditions to use AI without always depending on outside help. The package turns recurring decisions and tasks into editable playbooks, templates, rubrics, and guides with a defined audience, validation path, owner, and maintenance rule.

Next cases

PulsoAI news from four continents becomes a daily edition in Portuguese, with the source and the checks in plain viewAI Daily BriefingScattered news becomes a browsable audio briefing with a transcript and sources
A prototype applied to your context

Want to assess a guided prompting experience for your team?

I can walk through Lapidador's decisions and discuss a bounded prototype for a real workflow.

Discuss an AI prototype