DEV Community

Rohith Pavithran
Rohith Pavithran

Posted on • Edited on

StudyTrak // Astra Core – Gemma 4 Challenge Submission

#devchallenge #gemmachallenge #gemma

Gemma 4 Challenge: Build With Gemma 4 Submission

StudyTrak // Astra Core – Gemma 4 Challenge Submission

This is a submission for the Gemma 4 Challenge: Build with Gemma 4

What I Built

StudyTrak (Astra Core) is an immersive, high-fidelity student spaced repetition planner and neural coaching console. Designed to work completely local-first or proxy securely via remote endpoints, StudyTrak solves the "academic attention fragmentation" problem. It merges class schedules, homework tracker boards, spaced-repetition card decks (SM-2 scheduler), and deep neural academic helpers into a single HUD-inspired dashboard.

Core Features

  • 🗓️ Astra Chrono Scheduler: Interactive class blocks and room assignments synced with responsive real-time layout structures.
  • 🎯 Spaced Repetition Flashcards: Localized FSRS/SM-2 simulator (with grades: Forgot, Hard Link, Good Match, Perfect Spot) scheduling item intervals adaptively.
  • 🎙️ Multi-Modal Capture Panel: Scan worksheets (Camera/Photo uploads) or type natural language inputs to automatically parse schedule drafts and pending homework proposals.
  • 💬 Subject Advisory Chat Tutor: Academic conversational helper grounded directly in defined high-school/college subjects.
  • 🧘‍♀️ Empathetic Reflection Journal: Conversational weekly debriefs tracking achievements, struggles, and proposed behavioral corrections.
  • 🔊 Voice Speech Synthesis: Listen to high-fidelity audio versions of daily briefings with an interactive audio player.

Demo

Code

The complete, pristine full-stack source code is published on Gitea at https://gitea.com/rohithtp/studytrak. You can clone, audit, or deploy the application directly from this repository.

How I Used Gemma 4

StudyTrak leverages the lightweight yet highly competent capabilities of the Gemma 4 model lineup to deliver a privacy-preserving co-pilot:

  1. Model Choices and Fit:

    • Gemma 4 Tiny (E2B - 1.5GB): Configured as the optimal on-device WebLLM/MediaPipe companion. We chose the E2B formulation because it easily fits within the standard modern browser sandbox, enabling sub-150ms real-time token generation for micro-scans and speech transcription parsing without any outbound network latency.
    • Gemma 4 Standard (E4B - 3GB): Selected for core desktop semantic analysis, empathetic reflection summaries, and structuring homework proposals from handwritten worksheet snapshots. It balances rapid execution speeds with reasoning traits needed to provide high-quality structural mentorship.

  2. Context Grounding:
    Every prompt generated goes through a deep localized context packaging layer, supplying active subject indices, schedule constraints, and historic reflection streaks to completely ground prompt results with zero external leak risk.

Top comments (0)