kkg.dev
Engineering Portfolio

Case Studies

Deep-dives into backend architectures, performance metrics, and design schemas of production-ready systems.

Project Focus

AI-Powered Support Ticketing System

High-performance real-time customer support platform with automated resolution suggestions

An enterprise-grade support ticketing system supporting multi-tenancy. It dynamically processes incoming client requests, streams real-time updates to staff dashboards, and integrates Gemini AI to auto-draft suggested email replies.

Technologies Used

Node.js Express.js PostgreSQL Redis (Caching & WS Store) WebSockets (ws) Gemini AI API

Key Features

  • Real-time updates to agent queues via WebSocket connections, minimizing database polling.
  • Automated initial issue triage and resolution suggestions powered by Gemini Pro API.
  • Multi-tenant logical database isolation, allowing secure storage of proprietary ticket details per corporate account.
  • Granular Role-Based Access Control (RBAC) separating admins, managers, and support agents.

Performance Metrics

< 15ms

Socket Broadcast latency

3.4s avg

AI Resolution drafting

45ms (p95)

API Response Time

System Architecture Flow

1. Client Connection

A customer opens a ticket or logs in. A stateful WebSocket connection is established to the Express backend.

2. Task Queuing

Incoming ticket payload is validated and written to PostgreSQL. An event is dispatched to Redis Pub/Sub.

3. Gemini AI Analysis

A worker process reads the ticket and invokes the Gemini API to analyze sentiments and draft suggested answers.

4. Real-time Delivery

The drafted suggestions and ticket updates are pushed immediately to the agent console via WebSockets.

Project Focus

Credit Card Management System

Secure ledger database for managing credit card transactions and bill reminders

A financial platform built to help users manage credit card profiles, track daily transactions, calculate monthly outstanding billing statements, and configure payment reminders.

Technologies Used

Node.js Express.js PostgreSQL (Sequelize) Redis BullMQ Generative AI (Future Plan)

Key Features

  • Consolidated credit card ledger database storing card types, limits, statements, and transactions.
  • Automated cron job scheduling utilizing Redis and BullMQ to send statements and payment alerts.
  • Secure transaction logging with database transaction rollback checks, preventing race conditions.
  • Planned integration of an AI chatbot for context-aware financial query processing and expense analysis.

Performance Metrics

100% ACID

Ledger transaction audit

99.98%

Reminder dispatch reliability

12ms

Query execution time

System Architecture Flow

1. Transaction Log

User transactions are submitted and written to PostgreSQL inside a secure ACID transaction scope.

2. Cron Alert Engine

Daily scheduler scans upcoming payments and queues warning payloads in BullMQ.

3. Dispatch Queue

BullMQ workers process tasks and trigger notification gateways to send reminders.

Kundan Gupta