AI8 min
Teaching Defensive AI to Think Like a Red Team
A walkthrough of how I architected multi-agent simulations that pit autonomous attackers against defensive RL policies to harden enterprise networks.
Read article →
Ram Kansal
Autonomous security where reinforcement learning meets cyber warfare.
Command Palette
⌘K
~/omnis
›
Deploy
Scan
Train
Mitigate
Latency
23 ms
Policy
Zero-Trust
Status
Active
Journey
Building at the intersection
Highlights only — details live in Projects.
2025 – Present
OMNIS Defense System
Intent-aware security cloud that predicts and neutralizes threats before breach.
Reinforcement LearningThreat Intelligence
2024 – 2025
NASA Space Apps Challenge
Shortlisted finalist building adaptive AI for space infrastructure security.
Machine LearningLeadership
2023 – 2024
Offensive Framework Integration
Automated 100+ pentest tools into one framework; cut engagement time by 60%.
PythonAutomation
Featured Work
Systems that scale autonomously
From defensive AI to offensive automation — projects where research meets production deployment.
2025
active
pentestMCP
Autonomous MCP suite orchestrating reconnaissance, exploitation, and reporting from a single command surface.
Autonomous Playbooks
Configurable attack graphs that adapt based on live telemetry and surface new actions in real time.
Human-in-the-loop
Secure acknowledgement checkpoints ensure sensitive actions require explicit reviewer approval.
MCPGoTypeScript+3 more
CybersecurityAI
Open ↗2024
active
Adaptive Multi-Threat Mitigation
Deep reinforcement learning agents hardening networks by continuously sensing, simulating, and neutralising blended threats.
Adaptive Defenses
Policy updates streamed every 15 minutes based on live anomaly feedback and adversarial simulations.
Observability
Graph-based situational dashboards expose decision boundaries and confidence windows for every response action.
PyTorchRay RLlibGraph Neural Nets+2 more
AICybersecurity
2025
active
North-Star Exoplanet Classification
Full-stack ML platform triaging NASA Kepler signals into high-confidence exoplanet discoveries for Space Apps Challenge.
Explainability
Integrated SHAP overlays that justify every classification for mission review boards.
Collaborative
Realtime annotations let scientists interrogate light curves and leave decision trails.
Next.jsTypeScriptTensorFlow+2 more
AIFull Stack
Research & Insights
Knowledge from the frontlines
Deep dives into AI security, adversarial systems, and the future of autonomous defense.
6 minMCP
Designing pentestMCP for Autonomous Offensive Ops
Breaking down the architecture decisions that let pentestMCP compose reconnaissance, exploitation, and reporting without drowning operators in noise.
Read →
7 minIoT
Real-Time IoT Botnet Detection Using Graph Neural Networks
How graph neural networks can identify coordinated botnet behaviour across distributed IoT devices by modelling communication patterns as spatial-temporal graphs.
Read →
8 minAI
Explainable AI for Exoplanet Discovery: Lessons from NASA Space Apps
How we built North-Star, an interpretable ML pipeline that helps astronomers understand why a model flags a star system as hosting exoplanets.
Read →
9 minKubernetes
Zero-Trust Kubernetes: Hardening Cloud-Native Infrastructure
A practical guide to implementing zero-trust principles in Kubernetes clusters, from network policies to runtime threat detection.
Read →
6 minAI Security
Adversarial Attacks on Large Language Models: What Developers Need to Know
Exploring prompt injection, data poisoning, and model extraction attacks against LLMs—plus practical mitigations for production deployments.
Read →
Available for Collaboration
Let's build something extraordinary
Whether it's defensive AI systems, offensive security automation, or cutting-edge research collaboration — I'm always open to discussing ambitious projects that push boundaries.