Sakarey Parangi Bharath
I build and operate the systems security teams actually rely on: a home-built SOC that detects and triages real attack techniques, a fraud-scoring platform with live relay-attack detection, and an LLM vulnerability scanner that cut false positives 97% against a real production-scale repo. Currently looking for a cybersecurity internship.
Featured work
NFC Payment Simulator
Contactless payment stack — card emulator, POS terminal, token vault, ML fraud scoring, and a relay-attack timing detector with a live React dashboard.
SOC Home Lab
Virtualized SOC — Windows 10 endpoint + Ubuntu Splunk Enterprise — with simulated brute-force, phishing, and encoded PowerShell attacks triaged end-to-end and mapped to MITRE ATT&CK.
AI Security Auditor
CLI that scans GitHub repos for OWASP LLM Top 10 issues — hardcoded secrets, prompt injection, unsafe exec, insecure RAG. Tested on LangChain, cutting false positives ~97% (1,240 → 40).
Co-Evolutionary Red/Blue RL
A concept I'm circling, not yet built: attacker and defender agents training against each other inside a simulated vulnerable banking environment.
Rate Limiter Service
Token bucket rate limiter shipped as a real Flask API, stress-tested with a burst-traffic load generator — the classic Amazon system-design topic, built hands-on.
PhishGuard
Chrome extension that scores pages in real time for phishing risk using XGBoost / Random Forest models cross-checked against the PhishTank API.
Mini-SIEM
Lightweight log-ingestion SIEM with detection rules mapped to MITRE ATT&CK and automatic IOC extraction from raw logs.