Skip to content
View AnandSundar's full-sized avatar

Block or report AnandSundar

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
AnandSundar/README.md


⚑ What I Do

I engineer AI-powered security systems that think, detect, and respond β€” so humans don't have to chase every alert.

I sit at the intersection of Software Engineering, Agentic AI, Cybersecurity, Compliance Engineering, and Mainframe Development. My work focuses on four core areas:

πŸ€–  AI Security β€” Hardening LLMs and multi-agent systems against prompt injection, model poisoning, and adversarial attacks using the OWASP LLM Top 10 as the playbook.

πŸ›‘οΈ  GRC Automation β€” Writing the pipelines that replace manual audit busywork with continuous control monitoring, automated evidence harvesting, and real-time compliance drift detection β€” across NIST, SOC 2, ISO 27001, GDPR, HIPAA, PCI-DSS, and CMMC.

βš™οΈ  Security Engineering β€” Writing the automation that patches vulnerabilities, hunts threats, and locks down cloud environments β€” using Python, Go, PowerShell, and Bash.

πŸ–₯️  Mainframe Engineering β€” Programming IBM z/OS at the metal level β€” COBOL, Assembler, JCL, REXX, and Python on Z β€” on the platform that quietly processes over $10 trillion in transactions every day. While most engineers debate frameworks, I'm in the registers.


πŸš€ Project Showcase

πŸ€– Agentic AI Security

  • Agent Firewall - Default-deny stdio proxy for production AI agents, mediating MCP tool execution against strict SOC-action catalogs
  • Email Triage Agent β€” Inbox zero, finally β€” an autonomous email agent that handles the noise so you can focus on what matters.
  • TaxFlow AI β€” AI-powered tax document analysis that automatically processes, analyzes, and extracts insights from financial documents.
  • MindBridge AI Operations Hub β€” Automating the busywork of mental healthcare β€” so clinicians can focus on people, not paperwork.
  • Attack Surface Monitor β€” Real-time AI agent that continuously maps and scores attack surfaces before attackers find them
  • VaultHound β€” Autonomous credential and secrets leak hunter across git repos, CI/CD pipelines, and cloud configs
  • PhishSentinel β€” LLM-powered phishing detection engine with automated threat classification and triage
  • BountyOps β€” Agentic assistant that accelerates bug bounty operations with intelligent vulnerability triage
  • Cyber AI Log Analyst β€” AI agent that parses, correlates, and surfaces behavioral anomalies buried in security logs
  • Prompt Injection Scanner β€” OWASP LLM01:2025 β€” Automated red-team scanner for LLM prompt injection vulnerabilities

πŸ›‘οΈ GRC & Compliance Automation


βš”οΈ Threat Hunting & Vulnerability Management


πŸ“ Open Source Contributions

  • Axios PR #11029 - Network proxy bypass fix, canonicalized edge-case IPv4 formats (shorthand/octal/hex) to ensure robust no_proxy evaluation
  • opentag PR #58 - Populated owner/repo metadata to fix project-target binding lookup for GitLab integrations
  • FckSignups PR #213 - Added Simple Icons to enhance signup UI clarity and visual consistency
  • Agent Beacon PR #237 - Added 18 threat-detection rules across 6 categories, enhancing agent telemetry coverage and detection fidelity
  • AetherSDR PR #3709 - AppStream metadata refinement and release packaging, ensuring strict specification compliance for Flathub distribution
  • KAI OS PR #30 - 60-second local tour transcript, zero-install architecture mapping (Agent = Process, Tool = Syscall), markdown product proof rendering
  • GRCEngClub/claude-grc-engineering PR #67 - US-FINRA framework plugin stub, expanded AI-driven compliance architecture for financial regulatory standards
  • GRCEngClub/claude-grc-engineering PR #89 - HIPAA Security Rule framework plugin (Reference depth), expanding AI-driven compliance automation for healthcare data
  • wmux PR #280 - Agent-to-agent (A2A) channel architecture, implemented U1 domain types and underlying state persistence layer
  • vigil365 PR #2 - Per-alert snooze controls and silent auto-resolve logic for alerting workflows

Mainframe Programming - IBM Z/OS


Software Engineering

  • boris-loop - Starter kit for the loops over prompts pattern, distilled into ~100 lines of runnable Python
  • Offer Negotiation Simulator β€” Negotiate like a boss. Walk away with more.

🧰 Tools & Technologies

πŸ€– Agentic AI

πŸ” AI & LLM Security

πŸ›οΈ IAM & Policy-as-Code

βš™οΈ Languages & Automation

☁️ Cloud & DevSecOps

πŸ“‹ Governance & Compliance

πŸ” SIEM & Threat Detection

πŸ–₯️ Mainframe & IBM Z/OS


πŸ“ˆ GitHub Activity


contribution snake

🀳 Let's Connect

Pinned Loading

  1. go-idempotency go-idempotency Public

    Go 1

  2. React16-BurgerBuilder React16-BurgerBuilder Public

    A fun way to build and order burger online

    JavaScript

  3. Cyber-AI-Agent Cyber-AI-Agent Public

    Python