Current focus
AI and agentic systems
I threat-model agents that call tools, test LLM applications adversarially, and review the ML pipeline behind them, from training data to inference.
Bengaluru In security since 2013
Cybersecurity Architect & Researcher.
Adversarial testing and threat modeling for AI and agentic systems. Read the math, not the press release.
- Practice
- 13+ years
- Education
- M.Tech, BITS Pilani
- Sectors
- Defence · Banking · Retail · SaaS
Thirteen years in security. Application code first, then infrastructure, cloud and security programs, and now AI systems. Seven years of consulting across defence, power, manufacturing, telecom, retail, e-commerce and SaaS taught me to find the risk in an unfamiliar system fast; banking and insurance taught me what a wrong call costs. The Data Science M.Tech from BITS Pilani lets me check the ML math behind a security claim instead of taking it on trust.
ls ./practice
What I work on.
AI and agentic systems are the focus now. Everything else here is the 13+ years of work underneath it.
Offensive testing
Seven years of consulting: breaking web apps, APIs and mobile clients, then reading the source to see why they broke. My red-team work now covers LLM applications and agents too.
Architecture
Threat models and architecture reviews that settle where trust boundaries sit before anything ships. Cloud work across AWS, Azure and GCP: identity, segmentation, data and the CI/CD path.
Security programs
I built and ran vulnerability management for a regulated insurance broker: pipeline gates that catch problems without breaking the build, and findings that each have an owner and a deadline.
See the longer notes at /practice/
ls ./writing
More writing.
Threat modeling, red teaming, and where classical security thinking stops working for AI.
Red Teaming Generative AI: Language as the New Exploit Vector
Prompts are payloads. Why classical red-teaming misses LLM-native attacks, and how to design adversarial tests that surface jailbreaks and tool misuse.
Read articleIoT Devices Fail Open by Default: A Hardening Checklist
IoT devices fail open by default. A pragmatic checklist for hardening firmware, networks, and lifecycle management against the threats most teams overlook.
Read articleThe MAESTRO Framework for ML Security, Layer by Layer
CSA's MAESTRO framework, explained for practitioners. Layer-by-layer attack surface, control mapping, and how to apply it to your ML pipeline this week.
Read articleRead every post at /blog/