$ whoami
Karthik Bhaskar
Senior Applied Scientist at Vijil.ai
Building trust infrastructure for AI agents. Research in adversarial ML, LLM security, and trustworthy AI systems. Previously at CIBC, Vector Institute, University of Toronto.
recent posts
The OWASP Top 10 for LLMs: What Actually Matters
OWASP released its 2025 Top 10 for LLM Applications. Having built and secured LLM systems, here's which vulnerabilities actually bite and which ones are noise.
The GCG Attack: Three Years Later, We Still Haven't Solved It
In 2023, a single paper broke the safety alignment of every major LLM. Three years and dozens of defenses later, the core problem remains unsolved. Here's what happened.
Hello World: A New Beginning
Introducing my new website and what to expect from this blog - thoughts on trustworthy AI, adversarial ML, and building reliable AI systems.