Hi, I'm Gaanesh. I'm a security engineer in Singapore who breaks things on purpose, then figures out what happened.

Every DFIR investigation ends in a written report, and producing it is slow manual work that happens when an analyst is already tired. The open question was how much of an investigation an autonomous agent could reasonably carry.

• Built a self-reporting structure that drafts investigation write-ups from findings instead of leaving it to a human at the end.
• Researched autonomous agents for DFIR investigations: what can be delegated, where the reasoning breaks down, and which steps still need a person.

Internship work inside GovTech CSG. Implementation details stay internal.

Routine security questions meant a person digging through scattered internal documentation, and cloud security requests sat in a five-day queue.

• Built an internal RAG chatbot on an agentic architecture, so the system retrieves and reasons rather than keyword-matching.
• Automated the manual cloud security workflows behind the queue.
• Wired security checks into CI/CD through API integrations so problems surface at commit time.

Most online utilities for security work ask you to upload the very thing you are trying to keep private. I kept needing these tools and kept not trusting where the data went.

• Built a suite of utilities that run entirely in the browser: no uploads, no server, nothing leaves the tab.
• Self-hosted and open to anyone, so the privacy claim can be checked rather than taken on trust.