AI-Native Lesson Planning Workflow2024–Present

Sagini

AI Product Engineer (Solo, 3 months)

An AI-native workflow for generating structured lesson plans from teacher inputs using prompt design and guided outputs.

Next.jsOpenAIPineconeTailwind CSS

Problem

Teachers need faster ways to create structured, classroom-ready lesson plans without vague AI outputs.

Solution

Sagini uses guided inputs, prompt structure, and output templates to produce clearer lesson planning workflows.

Workflow

Teacher Input

Guided Prompt

Structured Output

Review

Lesson Plan

Key Features

Guided inputsStructured lesson plan outputsPrompt designAI-assisted workflowEducation-focused content generation

Workflow Architecture

Teacher InputGuided form

Prompt EngineTemplate + context

LLM APIStructured output

Lesson PlanCBC-aligned

Prompt Engineering & Structured Output

The biggest issue with LLM wrappers is unpredictable output. Sagini solves this by using highly constrained System Prompts and enforcing strict JSON schema parsing on the response. The application will not render the lesson plan unless the AI strictly conforms to the requested CBC curriculum format.

// Enforcing rigid structure on the LLM
const lessonPrompt = {
  system: "You are an elite CBC curriculum specialist. You must return your analysis wrapped in this exact JSON schema. Do not output markdown, pleasantries, or explanations.",
  schema: {
    subject: "string",
    grade: "number",
    strand: "string",
    learningOutcomes: "string[]",
    activities: "{ title, duration, methodology }[]",
    assessment: "string"
  }
};

// Usage in the edge function
const response = await openai.chat.completions.create({
  model: 'gpt-4-turbo',
  response_format: { type: 'json_object' },
  messages: [...]
});

Outcome

Generated 100+ CBC-aligned lesson plans during private alpha, significantly reducing teacher prep time.

Project Overview