Inspiration

Students in Bangladesh can find plenty of learning material, but choosing a realistic next step and maintaining a study routine can still be difficult. ShikhiAI focuses on the gap between a broad learning goal and a practical weekly action plan.

What it does

ShikhiAI creates a four-week learning roadmap based on a student's chosen skill, experience level, weekly study time, and learning challenge. Each milestone includes a practical deliverable. Learners can save multiple plans, check completed milestones, view overall progress, and download, export, or import a plan.

The five learning tracks are Web development, Data analysis, English communication, UI/UX design, and AI & prompt engineering. The interface supports English, Bangla, Arabic, Dutch, and German. The backend also supports AI translation of milestone text.

How it works

  1. Select a sample goal or enter learning preferences.
  2. Generate a personalized weekly roadmap.
  3. Complete practical tasks and mark milestones finished.
  4. Return to saved plans and use the Progress page to find the next unfinished task.

How it is built

The frontend uses HTML, CSS, and vanilla JavaScript, with browser storage for learner plans and completion checkmarks. A PHP backend validates requests and calls the Groq Chat Completions API using openai/gpt-oss-120b. AI output must contain exactly four milestone strings before it is accepted. The API key stays on the server, outside the public web root or in a server environment variable.

Request budgets use locked counters outside the public web root. If AI is unavailable, the app provides a clearly labeled fallback study plan rather than presenting it as AI-generated.

Technical challenges and lessons

The implementation addresses unreliable model responses through structured JSON requests and response validation. It separates browser-local learner records from AI requests, supports multilingual layouts including right-to-left text, and aggregates progress across saved plans. These choices highlight the importance of reliable state management and clear AI disclosures in a usable learning tool.

What makes it useful

ShikhiAI combines personalized planning, concrete deliverables, and a simple follow-through workflow without requiring a learner account. The demo demonstrates Data analysis and React/backend learning goals, saved-plan management, milestone tracking, export/import, and language switching.

Privacy and limitations

Names, saved plans, and checkmarks remain in the current browser. AI generation sends the selected goal, level, weekly hours, challenge, and language through the server to Groq; translation sends milestone text and the requested language. Users should avoid sensitive information. Progress reflects completed milestones, not verified skill mastery. Roadmaps are suggestions, not guarantees of educational or employment outcomes.

What's next

Potential improvements include curated learning resources, more detailed time estimates, and adaptive replanning based on learner feedback.

Links

Built With

Share this project:

Updates

Submission history