Pravinraj

Auto Scaling Under Live Load

AWS Infrastructure  ·  Load Testing  ·  May 2026  ·  LinkedIn Post →

• Simulated a DoS-style load test using Apache Bench — 10,000 requests at a concurrency of 100 — against an internet-facing Application Load Balancer.
• Configured a CloudWatch Target Tracking policy capped at 50 req/target, successfully triggering the ASG to scale from 1 to 3 instances under a peak load of 594 req/target.
• Diagnosed an initial 100% connection timeout rate to a Security Group misconfiguration, and pushed the architecture to the AWS account ceiling, triggering a 16 vCPU quota limit.

Incident Sentinel

AWS Serverless Security Pipeline  ·  Dec 2025  ·  View on GitHub →

• Architected a 100% serverless, event-driven security pipeline using S3, Lambda, and SNS — eliminating idle compute entirely and reducing costs to $0.
• Integrated Amazon Rekognition for real-time image analysis, achieving sub-second threat detection and triggering automated email alerts within <2 seconds of an S3 upload.
• Enforced strict least-privilege IAM policies across 3 separate AWS service boundaries to ensure complete secure isolation.

AI Lead Generation Chatbot 1st Place · Datronix Symposium

SRM Easwari Engineering College  ·  2-person team  ·  Feb 2026

• Built a domain-aware AI chatbot for lead generation using Python and AnythingLLM, winning 1st place among competing teams.
• Designed the backend to capture and persist conversation leads into a SQLite database via a REST API.

Magic Until You Read the Spec

Blog  ·  May 3, 2026  ·  Read Article →

I spent years thinking AI was something close to magic. Then I started building with it — and the curtain fell hard.