Gojo
Hybrid AI Phishing Detection System
Overview
A production-grade phishing URL detection system combining heuristic rule-based analysis, ensemble machine learning, and reinforcement learning. The system leverages lexical and character n-gram feature models alongside a Thompson Sampling-based contextual bandit for adaptive real-time learning and decision-making. Built with a multi-layer detection architecture (feature extraction, content inspection, drift detection), it supports both real-time analysis with ultra-low latency and bulk processing. It is exposed via REST API, CLI, and web interface, aligning with practical SOC-style triage workflows.
"Gojo achieves ~99% accuracy by marrying traditional heuristic rules with advanced ensemble ML and reinforcement learning, delivering SOC-ready performance."
Key Metrics
~99%
Detection Accuracy
~50K URLs
Dataset Size
~20ms
Latency
Technologies
Key Features
Hybrid Detection Engine
Combines heuristic rule-based checks with machine learning to identify complex phishing patterns efficiently.
Adaptive Learning
Integrates Thompson Sampling contextual bandits for real-time adaptation and decision-making against evolving threats.
Scalable Architecture
Engineered to support both real-time low-latency checks and bulk analysis (10,000+ URLs/batch) via API and CLI.
Open to Opportunities
Available for remote, hybrid, and on-site roles in security analysis, SOC, and systems engineering.
Cybersecurity · Systems Engineering · Security Automation
Security-first mindset, systems-level execution.
Final-year engineering student with practical experience in VAPT, phishing detection, automation tooling, and embedded control systems. Open for internships and full-time opportunities.
Get In Touch
Let's Connect
Open to internships, full-time roles, and project collaboration in security and systems.
Currently available
Open to internships, graduate roles, and freelance security or systems projects.