Skip to main content
Independent product · Shopnopath

A learning platform that follows the whole attempt.

From a photographed Bangla answer sheet to timed IELTS/SAT practice, I built the workflows that connect content, assessment, feedback, and paid access.

FastAPINext.js 16React 19PostgreSQLCelery
Shopnopath - AI Learning & Test Preparation conceptual project illustration

ROLE

Solo full-stack developer

2026 · Ongoing product

PRODUCT CONTEXT

Shopnopath began as a job-exam platform. Its latest work adds IELTS/SAT preparation and course authoring alongside the original job-exam platform.

Engineering contributions

What I built, and why it matters.

01

The server owns the exam clock

Attempt deadlines, autosave, and expiry recovery keep exam timing tied to stored state. Closing a tab does not stop an attempt, and background jobs can finalize expired work.

  • Live, model, and practice exams
  • Negative marking and randomized question pools
  • Autosave retries and tab-close answer flushing
02

Paper answers become reviewable work

Candidates can upload photographs of handwritten answers. The backend optimizes and stores them privately, queues transcription, and feeds typed text plus transcripts into rubric-based grading.

  • Visible OCR state and stuck-job recovery
  • Original page beside its transcript for moderators
  • Correction and regrading; failed OCR goes to review
03

IELTS and SAT get their own engine

A separate section-state engine supports computer-delivered IELTS practice and adaptive SAT mocks. It shares platform services while keeping its timing, routing, and score reports separate from the job-exam engine.

  • Server-owned section clocks and stored SAT routes
  • Configurable score tables and versioned reports
  • Listening, Writing, Speaking, and calculator experiences
04

Feedback with a recovery path

Writing and Speaking evaluation returns criterion feedback and examples. Automatic results have explicit retry and review rules; human corrections become current without discarding the AI history.

  • Multiple Writing runs with disagreement handling
  • Speaking feedback from audio with estimated pronunciation
  • Moderator escalation and student result-reporting flows
05

A product around the assessment

Courses, authoring tools, moderation, subscriptions, and checkout make the learning workflow usable beyond a single test. Generated content stays in draft until approved; language-test evaluation has separate automatic-release rules.

  • Course/test JSON import and export
  • bKash and SSLCommerz payment workflows
  • Next.js web and Flutter clients

IELTS/SAT results are practice estimates. This project does not claim official test affiliation or guaranteed exam scores.