CV Optimizer Backend

Backend service for an AI-powered CV optimization platform that helps users tailor their resume and generate cover letters for specific job descriptions using large language models.

Users upload their CV and a job description, and the system analyzes the content, generates an optimized version of the CV, and produces a targeted cover letter while tracking AI usage through a credit system.

This repository demonstrates backend engineering depth across architecture, domain modeling, reliability, security, and operational concerns, not just endpoint wiring.

Highlights

• Modular NestJS architecture with clear bounded contexts (auth, users, credits, files, optimizations, llm, email)
• Database-first approach with explicit SQL migrations and TypeORM entities (synchronize: false)
• Credit-ledger accounting model with idempotency support for safe, auditable balance changes
• JWT + refresh-token auth flow with OTP-based login
• File ingestion pipeline for PDF, DOCX, TXT, and MD
• LLM orchestration layer with provider abstraction (OpenAI integration designed to remain vendor-agnostic)
• Structured validation, error handling, and configuration layering for production readiness

Tech Stack

• Runtime: Node.js, TypeScript
• Framework: NestJS 11
• Database: MySQL + TypeORM
• Auth: JWT + Passport
• AI: OpenAI Node SDK
• Email: Nodemailer
• Testing: Jest + Supertest

Architecture

The backend follows a layered architecture typical of production NestJS services.

Request flow:

1. Controller receives request and validates DTO input.
2. Global JwtAuthGuard enforces authentication by default (with explicit @Public() exceptions).
3. Service layer executes domain logic.
4. Repositories/entities persist state changes in MySQL.
5. Cross-cutting concerns (global exception filter, validation pipe, config service) apply consistently.

Key structural decisions:

• Module isolation: each domain owns its controller/service/models.
• Manual migrations: avoids drift and accidental schema mutations.
• Ledger-based credits: immutable transaction history for traceability.
• Provider abstraction for LLM: business logic is decoupled from vendor SDK details.

Domain Model Snapshot

The system is designed around a credit-based AI usage model where LLM operations are metered and auditable.

Core entities include:

• users
• otp_codes
• refresh_tokens
• optimizations
• credit_ledger
• llm_usage
• emails_sent

This model supports secure auth, metered AI usage, and auditability of user credits.

API Surface (v1)

Base path: /api/v1

• POST /auth/send-otp
• POST /auth/verify-otp
• POST /auth/refresh
• POST /auth/logout
• GET /users/me
• GET /users/stats
• GET /credits/balance
• GET /credits/history
• POST /credits/add
• POST /files/upload
• GET /files/:filename
• POST /optimizations
• GET /optimizations
• GET /optimizations/:id

Local Setup

1. Install dependencies

npm install

2. Configure environment

cp .env.example .env

Set required values in .env:

• DB_HOST, DB_PORT, DB_USERNAME, DB_PASSWORD, DB_DATABASE
• JWT_SECRET
• EMAIL_HOST, EMAIL_PORT, EMAIL_USER, EMAIL_PASSWORD, EMAIL_FROM
• OPENAI_API_KEY

3. Run database migrations

mysql -u root -p cvoptimizer < src/database/migrations/001_initial_schema.sql
mysql -u root -p cvoptimizer < src/database/migrations/002_add_input_file_format.sql

4. Start development server

npm run start:dev

Server: http://localhost:3000/api/v1

Quality and Security Notes

• Global input validation (whitelist, forbidNonWhitelisted, transform)
• Global exception filtering for predictable API errors
• Environment-based configuration via Nest ConfigModule
• .env and uploaded user files are git-ignored
• Production deployment should use managed secrets and rotated credentials

Repository Docs

Project and implementation documents are in Docs/.

What This Repository Demonstrates

• Designing maintainable backend services with clear modular boundaries
• Modeling transactional business logic with auditability requirements
• Integrating LLM-powered features while controlling usage and cost
• Building production-ready APIs with migrations, validation, and structured error handling

Future Improvements

Possible extensions to this system include:

• background job processing for long-running optimizations
• streaming LLM responses
• vector search for resume/job similarity scoring
• usage analytics and cost monitoring dashboards