Full Description

Project Overview

Developed during my tenure at a previous organization, this enterprise application focuses on delivering a high-throughput, multi-modal assessment engine supporting thousands of active concurrent production users. It seamlessly processes multiple testing modalities—including standard MCQs, structured aptitude tests, complex coding environments, and subjective question sets—while generating granular performance analytics and AI-driven candidate insights.

Key Contributions & Technical Architecture

Core Architecture: Collaborated with a backend engineering squad to architect the system from scratch, defining core relational data models and clear REST API contracts for the end-to-end evaluation lifecycle.
AI Evaluation Engine: Integrated the Gemini API (Vertex AI) to architect an automated subjective evaluation engine and a conversational AI interview module, removing human grading bottlenecks and delivering real-time candidate feedback.
Sandboxed Code Execution: Integrated safe third-party execution APIs (Judge0) to handle isolated, real-time code execution and scoring for automated student programming evaluations.
AI-Driven Automated Scoring: Designed custom prompt engineering patterns and backend workflows that utilize LLMs to analyze code logic, architectural choices, and subjective text inputs against robust baseline rubrics.
Database Optimization: Eliminated $N+1$ query overheads using targeted fetch joins and optimized stored procedures, accelerating reporting and analytical dashboard loads from 12 seconds to under 2 seconds.
Asynchronous Processing: Leveraged Spring @Async annotation primitives for resource-heavy code evaluation processing and AI orchestration, keeping main API threads responsive under heavy concurrent loads.
Data Integrity: Wrapped complex multi-step write processes using @Transactional boundaries to prevent partial database failures and maintain data synchronization across independent tenant nodes.
Analytical Dashboards: Built comprehensive reporting modules tracking institutional metrics such as aggregate passing percentages, average scores, time efficiency, and topic-wise skill gaps.
Design Patterns: Refactored a legacy authentication filter and multi-tier role routing layout using the Factory Design Pattern, bringing a messy code layer into an easily extensible, highly readable security module.
Multi-Tenancy: Managed features within a multi-tenant monolithic backend structure designed to serve multiple corporate client portals from a single, resource-efficient deployment instance.
Observability & CI/CD: Set up health metrics via Spring Boot Actuator, established structured logging frameworks (SLF4J/Logback) to speed up live production debugging, and implemented basic automated testing workflows via GitHub Actions.

Tech Stack Used

Backend Frameworks: Java, Spring Boot, Spring Data JPA, Hibernate, Spring Security, REST APIs
AI & Isolation Runtimes: Gemini API (Vertex AI), Judge0 APIs
Database & Security: PostgreSQL, JWT, RBAC, Transaction Management
Observability & DevOps: Spring Boot Actuator, SLF4J, Logback, Docker, GitHub Actions

PrepEngine

Highlights

Key Features