01
Open to Opportunities

VARUN
RATHOD

CS student at PES University, Bangalore. I build AI agents, security tools, and full-stack systems — from Linux kernel modules to Deep RL game agents.

CTF competitor · Hackathon finalist · Club Head

50K+
Hangman Corpus
100+
Hackathon Participants
02
Club Roles
02
Introduction

About Me

I'm a Computer Science student at PES University, Bangalore, with a habit of building things that sit at the intersection of systems, intelligence, and security.

My projects range from Linux kernel-level systems programming to training Deep RL agents on raw pixel frames. I compete in CTFs, have helped organize a 100+ participant hackathon, and build production-grade full-stack apps as a way to close the loop between theory and real use.

I care about understanding how things work at a low level — whether that's a process tree in the kernel or the emission matrix of an HMM.

Cybersecurity

AI / ML

Systems

Full-Stack

03
Technical Expertise

Skills

Languages

01
Python
C
JavaScript
Java
R

Frameworks

02
React
Node.js
Express
Next.js
Flutter

Databases

03
MongoDB
PostgreSQL
MySQL
Neo4j
Redis

AI/ML

04
PyTorch
Deep RL
Generative AI
OpenCV

Security

05
Cryptography
Web Security
Network Security
OSINT
CTFs

Tools

06
Linux
Git
Docker
Socket Programming
Burp Suite
04
Selected Work

Projects

A curated collection of my best work in AI, security, and full-stack development.

01

Second Brain — AI Knowledge System

An AI-powered knowledge management system where you can dump notes, documents, and ideas, then query them conversationally. Built with Next.js, Prisma, PostgreSQL, and Gemini API. Deployed on Vercel.

Full auth flow with session management
Organizes notes into queryable knowledge graph, not just keyword search
Semantic summarization via Gemini API
Production deployment on Vercel
Tech Stack
Next.jsPrismaPostgreSQLGemini APIVercel
02

Hangman AI Agent

A hybrid AI that plays Hangman by combining Hidden Markov Models with Q-Learning. Trained on a 50,000-word corpus and evaluated on 2,000 completely unseen words. Achieves 100% win rate on 20+ letter words.

Adaptive 6-strategy weighting system
Shifts from frequency-based to pattern-based reasoning
100% win rate on 20+ letter words
Trained on 50,000-word corpus
Tech Stack
PythonNumPyHMMQ-Learning
03

Slither.io Deep RL Agent

Trained a CNN-based DQN to play Slither.io from raw pixel frames alone — no game API, no coordinates, no snake positions. Just pixels in, actions out. Built a custom Gymnasium environment, implemented replay memory and epsilon-greedy exploration.

Pure pixel input → policy, no hand-engineered features
Custom Gymnasium environment for game simulation
Experience replay buffer with prioritized sampling
CNN-based DQN with target network stabilization
Tech Stack
PyTorchOpenCVGymnasiumDeep RL
04

Traffic Violation Detection System

Full-stack web application for managing traffic violations, with MySQL triggers and stored procedures that auto-suspend licenses when point thresholds are hit. REST API with connection pooling + React frontend.

Database-driven business logic
Triggers fire on violation inserts to calculate point totals
Automatic suspension eligibility evaluation
Connection pooling for performance
Tech Stack
Node.jsExpressReactMySQLREST API
05

VoIP with Progressive Security

Full-duplex voice communication system built over raw sockets with layered security: starts unencrypted, progresses to TLS with mutual certificate authentication. Built entirely from scratch — no VoIP libraries.

Mutual TLS with custom certificate authority
Handshake negotiation built at application layer
Upgradeable security: plaintext → TLS without restarting
Full-duplex audio over raw TCP sockets
Tech Stack
PythonSSL/TLSRaw SocketsCryptography
06

Web Scraper Price Comparison

Async price scraper across 5 e-commerce platforms with proxy rotation and an analytics dashboard. Handles anti-bot measures, rate limiting, and parallel scraping without getting blocked.

Async scraping across 5 platforms
Proxy rotation to avoid blocking
Handles anti-bot measures
Analytics dashboard
Tech Stack
FlaskReactSeleniumPython
07

Process Tree Kernel Module

A Linux kernel module written in C that traverses the process hierarchy using the task_struct linked list and outputs the full tree to the kernel log. Loaded and unloaded dynamically via insmod/rmmod.

Traverses task_struct doubly-linked list in kernel space
Outputs full process hierarchy to dmesg / kernel log
Dynamically loadable via insmod / rmmod
Zero userspace dependencies — pure kernel module
Tech Stack
CLinux KernelSystems Programming
05
Recognition

Achievements

01

PBCTF 4.0

Top 15web + crypto challenges

2025
02

Codefusion Hackathon

4th Place24h build

2024
03

Kalpana Hackathon

Top 10

2025
04

niteCTF

~350 / 800 teams

2025
05

OverTheWire: Bandit

Completed all levels

2025
06
Leadership

Clubs

Equinox ECC

Club Head

Organized a 16-hour Hackathon + CTF with 100+ participants. Handled end-to-end logistics: problem setting, judging, and running the event live.

Layer8

Tech Member

Author of CTF challenges used in the Layer8 × IEEE competition. Conducts workshops on cybersecurity topics including web security, binary exploitation, and more.

07
Get In Touch

Contact

Open to collaborations, opportunities, and interesting conversations.