Welcome to my GitHub.

I'm an undergraduate Computer Science and Engineering student passionate about building secure, scalable systems and exploring machine learning, system design, and backend development.

I enjoy solving real-world problems with thoughtful engineering and clean, efficient code.

• Backend Development & System Design
• Applied Machine Learning & Natural Language Processing
• Cryptography & Blockchain Concepts
• Distributed Systems & Cloud Infrastructure (Learning)
• DevOps (Docker, Kubernetes, Jenkins, NGINX etc) (Learning)
• Competitive Programming & Algorithms

A collection of production-grade systems, and large-scale backend applications focused on distributed systems, real-time data processing, security, and infrastructure engineering.

Most of these systems are deployed in real environments, built with an emphasis on correctness, scalability, observability, and clean system design.

I have heavily been involved with the entire system design along with participating heavuly in backend developemt :)

Official deployment for SSN College of Engineering & Shiv Nadar University Chennai
Production system serving students and faculty | 2025

🔗 Live System: http://bustracker.snuchennai.edu.in/

A high-performance real-time mobility tracking backend designed to support thousands of concurrent users during peak campus commute hours.

• 3,000+ concurrent users supported during peak traffic hours
• Sub-50ms live location propagation from ingestion → map render
• Redis-backed in-memory state engine for millisecond reads
• Hot-path isolation removing databases from live request paths
• Concurrent GPS ingestion from multiple heterogeneous vendor APIs
• Adaptive polling system based on data freshness and load
• Full on-premise deployment with production observability stack
• Concurrent GPS ingestion: ~20 vendor GPS APIs are polled in parallel using Java threads (ExecutorService/CommandLineRunner), with time-of-day adaptive polling intervals to prevent cascading delays if a provider responds slowly.

                 +----------------------+
                 |   GPS Hardware       |
                 | (Multiple Vendors)   |
                 +----------+-----------+
                            |
                            |
                     GPS API Polling
                            |
                            v
                +-----------------------+
                |   Spring Boot Backend |
                |  (Location Processor) |
                +-----------+-----------+
                            |
                Normalize / Clean GPS Data
                            |
                            v
                 +----------------------+
                 |        Redis         |
                 | In-Memory Live State |
                 | busId → GPS location |
                 | Caches student bus   |
                 | allocation and exam/ |
                 |  breakdown overrides |
                 +----------+-----------+
                            |
             +--------------+--------------+
             |                             |
             v                             v
     +---------------+            +---------------+
     |  Mobile App   |            |  Web Client   |
     | (Students)    |            | (Tracking UI) |
     +---------------+            +---------------+

         Monitoring & Observability Layer
      Prometheus → Metrics → Grafana Dashboards

The system is engineered to support thousands of simultaneous users during campus rush hours.

• Redis Sets acting as a high-speed in-memory state engine
• No database access in the live request path
• Read-heavy endpoints optimized for in-memory access

• <50ms location propagation latency
• Smooth UI updates during morning and evening traffic spikes
• Stable performance during high concurrency bursts

Campus buses use different GPS hardware vendors, producing inconsistent and noisy data streams.

The Spring Boot backend acts as a normalization layer, responsible for:

• Aggregating GPS signals from heterogeneous vendor APIs
• Cleaning noisy or inconsistent location updates
• Smoothing position data before publishing to clients
• Converting all GPS inputs into a single unified location format

The backend currently exposes ~60 API endpoints used by mobile and web applications.

Campus transport usage spikes dramatically during commute windows.

To handle this load, the platform uses:

• Containerized services for predictable deployments
• Each GPS API (~20 individual calls) are done using a seperate Java Thread using CommandLine Runner or Executor service with dynamic sleep times based on the time of the day for no cascading delays incase one api takes time for responding.
• Hot-path optimization eliminating database reads
• Adaptive GPS polling intervals to reduce unnecessary upstream requests

These techniques ensure the system remains stable even when thousands of users open the app simultaneously.

The system runs entirely on university infrastructure with full operational visibility.

• Prometheus → system and application metrics
• Grafana → real-time operational dashboards
• Custom alert pipelines

Operational alerts allow issues to be detected and resolved within minutes, often before users notice service disruption.

• Java
• Spring Boot
• Redis
• PostgreSQL
• Docker
• NGINX
• Prometheus
• Grafana

Deployed for SSN Invente Technical Fest | 2025

A Kafka-based real-time payment reconciliation platform built to process thousands of digital transactions with strong correctness guarantees.

• Event-driven reconciliation pipeline using Kafka
• Stateful stream processing with joins, aggregations, and windowing
• Redis-backed email ingestion system for payment receipt detection
• Idempotent processing to prevent duplicate reconciliation
• Load-tested at 50,000+ records per minute using K6
• Fully containerized production deployment

• Kafka producers and consumers
• Kafka Streams with time-windowed joins
• Redis-backed deduplication
• Transaction reconciliation pipelines
• Observability and throughput benchmarking

Tech Stack
Java • Spring Boot • Apache Kafka • Kafka Streams • Redis • PostgreSQL • Docker • K6

🔗 https://github.com/AvGeeky/kpaymentpipeline

Deployed for SSN College | 2025

A production-grade attendance system designed to eliminate proxy marking and manual verification overhead.

• Dynamic time-rotating QR codes
• One-time attendance codes for quick sessions
• Manual mode with teacher overrides
• Cryptographic device-bound verification
• Live attendance dashboards
• On-prem deployment

• HMAC-based authentication using device-bound secrets
• Time-limited code expiry
• Replay attack prevention
• Redis-managed verification windows

• Redis for real-time verification state
• MongoDB for persistent academic records
• Time-based validation engine
• SSE-driven live updates

Tech Stack
Java • Spring Boot • Redis • MongoDB • Docker • NGINX • JUnit • Jenkins

🔗 https://github.com/AvGeeky/AttendanceSystemBE 🔗 Demo: https://www.youtube.com/watch?v=YFHxfV337GY

A secure digital voting system focused on voter authenticity, privacy, and transparent verification.

• RSA-based vote encryption
• HMAC-secured Magic IDs for vote uniqueness
• PBKDF2-based authentication
• OTP login flow
• Immutable blockchain-style vote ledger
• Public vote verification interface

• Two-phase cryptographic voting
• Anonymous ballot storage
• Tamper-evident ledger design
• Modular backend services

Tech Stack
Java • Spring Boot • MongoDB • Cryptography • Docker

🏆 1st Prize — Encender Project Expo (CSE Department)

🔗

• Website: https://securvote.vercel.app
• Demo Video: https://youtu.be/4nAzh8N8cbA
• About: https://securvote.vercel.app/about

An intelligent interview platform combining technical assessment with behavioral analysis.

• Resume- and rubric-driven question generation
• Live coding interviews via Judge0
• Face detection and audio processing
• Sentiment analysis and automated summaries

🏆 1st Place — InnovateX National Hackathon (100+ teams)

Tech Stack
Python • AI/ML • Judge0 • Audio Processing

A collaborative learning ecosystem focused on incentivized education.

• Gamified in-app currency system (Stars)
• Community-generated learning roadmaps
• Dynamic teacher–learner role switching

Built within 24 hours during Envision SSN.

Tech Stack
Spring Boot • Web • Gamification Systems

• 🚆 Railway Reservation System (C) — Seat allocation, scheduling, and ticket workflows
• 🧍 Missing Persons Identification System — Facial recognition–based detection platform (high school project)

C	Java	Python	SQL

Spring	Flask	Scikit-learn

MongoDB	PostgreSQL	Redis	Docker	AWS

Git	GitHub	Linux	JetBrains	MongoDB Compass	Postman	Figma	Canva

• CLEF 2024 (JOKER Lab) – Genre classification of humorous text using transformer-based and traditional ML models.
• ILSUM 2024 – Multilingual text summarization using advanced NLG models like T5 and Gemini 1.0 Pro.