Hello, I'm Akhil Sharma — a polyglot engineer obsessed with AI security, LLM red teaming, and building agentic systems that are actually safe. I work across Rust, Go, Haskell, Python, Elixir — and spend my time jailbreaking AI models and figuring out how to stop others from doing the same.

Most recently, I was Chief Scientist & Co-Founder at SnowCrash Labs, an AI alignment startup where I built the core red-teaming evaluation framework and misalignment detection engine from scratch.

Before that, I founded ARMUR AI, backed by Techstars and Outlier Ventures, selected for Google AI Startups & Nvidia Inception. We built AI agents that scan codebases for vulnerabilities — and secured $500K+ in grants from Aptos, Tezos, and Algorand.

I'm a published researcher — my paper XAMT: Bilevel Optimization for Covert Memory Tampering in Heterogeneous Multi-Agent Architectures is on arXiv (2512.15790), formalizing adversarial attacks on multi-agent systems. I'm also a published author ('Rust for Blockchain Application Development' via Packt Publishing), and a LinkedIn Learning instructor on AI cybersecurity.

I teach 52K+ subscribers on YouTube at AkhilSharmaTech and share open-source security tools on GitHub — including Pentest-Swarm-AI (201 stars) and vibescan.

I teach 52K+ subscribers on YouTube at AkhilSharmaTech and share open-source security tools on GitHub — including Pentest-Swarm-AI (201 stars) and vibescan.
Stanford & MIT trained. AWS & GCP certified. 3x founder. Currently open to remote US roles in agentic security, LLM red teaming, and AI security tooling.