Open to research collaborations

I build things

that think.

> _

AI Researcher at SISA Information Security. B.Tech Cybersecurity, MIT Bengaluru '26. Building threat models for AI systems and automating security consulting.

3+
years in cybersec
4
repos shipped
'26
MIT grad
90/90
tests passing

About

I work at the intersection of AI and offensive security — building threat models for AI systems, researching adversarial ML, and helping organisations understand what breaks before attackers do.

My edge: I do real pre-sales consulting while still studying. That means I understand how security decisions are made commercially, not just technically. I build tools that collapse the gap between research and delivery.

2026
AI Security Researcher
SISA Information Security
Adversarial ML research, LLM security analysis, AI threat modelling for enterprise clients.
2022–
B.Tech Cybersecurity
MIT Bengaluru
Security testing, network forensics, applied cryptography, incident response.
2024–
Pre-Sales Security Consultant
SISA
Scope pen tests, write proposals, run discovery for ISO 27001, PCI DSS v4.0, SOC 2, HIPAA, DFIR engagements.

Skills

AI Security
Adversarial MLLLM Threat ModellingAI Zero-Day ResearchPrompt Injection AnalysisModel Extraction
Security Testing
OWASP ZAPWeb App PentestingNetwork PentestingAPI SecurityCloud SecurityVA&M
DFIR
Incident ResponseDigital ForensicsThreat HuntingTabletop ExercisesForensic Analysis
Compliance
PCI DSS v4.0ISO/IEC 27001SOC 2HIPAAGDPR
Engineering
React / Next.jsNode.js / ExpressPythonSQLiteClaude APIVercel AI SDK
Pre-Sales
Discovery & ScopingProposal WritingEffort EstimationDemo ScriptsPipeline Automation

Shipped Projects

proposal-engine

live
Agentic pre-sales automation — 6 BUs, 22 services

Production multi-agent pipeline that automates the full pre-sales cycle: discovery questionnaire, scoping, pricing, and proposal generation. Built on Groq (Llama 3), zero-cost deployment on Streamlit Community Cloud. Dark web3 UI with real-time pipeline progress.

PythonGroqLlama 3.3-70bStreamlitPyYAMLopenpyxl
5-agent pipeline: Questionnaire → Discovery → Scoping → Pricing → Proposal
22 service codes across 6 business units with codebook-driven pricing
90/90 tests passing — zero API calls in CI
Deployed: Streamlit Community Cloud (free, auto-deploy from GitHub)

presales-automation

live
AI-powered DFIR pre-sales pipeline

Full-stack tool that automates the entire pre-sales workflow for a DFIR security firm. Opportunity tracking, AI-generated DOCX proposals (IFI/Retainer/BAS), Claude-powered follow-up drafting, GAM contacts, executive dashboard with charts.

ReactViteExpressSQLiteClaude APIbetter-sqlite3
Claude API with prompt caching (cache_control ephemeral)
3 DOCX proposal types generated programmatically
Security-hardened: helmet, rate limiting, input validation
Excel import/export for bulk opportunity management

claude-code-best-practices

live
Production patterns for Claude Code

Reference guide built from real engineering work — not tutorials. Covers CLAUDE.md setup, prompt patterns with file:line references, security non-negotiables for backend code, token efficiency, and agent usage. Includes a working README generator CLI.

Node.jsClaude APIVercel AI SDKMarkdown
README generator: JSON config → Claude → GitHub-ready profile
Token efficiency patterns (prompt caching, maxTokens per function)
Security rules: SQL parameterization, enum validation, path traversal
CLAUDE.md templates for project-level and global conventions
SECOND BRAIN

Knowledge Graph

My Obsidian vault — auto-synced to GitHub after every session. Graph topology only. No note content exposed.

0
NOTES
0
CONNECTIONS
LIVE
SYNCED
root
journal
research
AUTO-SYNCED

Powered by Obsidian · Synced via second-brain repo

Ask Me Anything

Ask about my work, research, or how to collaborate. Pick your model — all stream via Vercel AI SDK Edge Runtime.

💬
Powered by Groq Llama 3.3 · Free · 300 tok/s

Chain: Groq (primary, free) → OpenRouter (fallback, free) · Vercel Edge Runtime · auto-failover