Welcome to Vani's GitHub

< / >

"Sic mundus creatus est."

---

Hi, I'm Vani, an Information Science student who's genuinely passionate about CS, not just as a degree but as a subject. I'm driven by the need to build, the hunger to learn, and the obsession with getting better at both. Cool projects, real problems, and a lot of trial and error, that's where I live.

• 🔭 Currently building in the AI security and LLM eval space
• 🛡️ Researching prompt injection, agentic betrayal, and adversarial testing
• 🌱 GSoC 2026 applicant — contributed to Catrobat/Catroid as part of the application
• 📍 Bengaluru, India

• 👾 WRAITH ── RL-based adaptive boss AI trained with GRPO on Qwen2.5-1.5B. Enemy that actually learns your patterns.
• 🛡️ AEGIS ── Prompt injection & agentic betrayal benchmark. Simulates poisoned DevOps environments to stress-test AI agent integrity.
• 🚀 ELEVATE ── AI-driven placement prep platform with adaptive skill assessment, voice confidence analysis, and sprint generation.
• 📊 SPECTRA ── Unit test analysis tool with live WebSocket terminal output, coverage dashboards, and visual test reporting.

• META × SCALER OPENENV HACKATHON ── Built and deployed WRAITH in 48hrs
• IEEE NEXTGEN CONVERGENCE HACKATHON 2026
• CODE DETECTIVE : BUG BOUNTY 2026

• AI / ML: Python FastAPI PyTorch LangChain GRPO RAG
• Security: Prompt Injection LLM Red Teaming Adversarial Eval
• Web: React Next.js Node.js MongoDB