Nilax Modi

I’m an AI-security and software development intern who builds secure, LLM-powered systems end-to-end—from architecture to encryption workflows to red-teaming. I enjoy designing components from zero and turning security requirements into measurable performance.

At SAC, ISRO, I’m architecting a dual-mode P2P secure chat application using PyQt6/Python. I support both QKD-based and Diffie-Hellman + AES-256 key exchange, achieving sub-200ms key retrieval by using direct UART-based provisioning from a Key Manager and reducing key exchange latency by 70% versus network-dependent approaches.

I also designed session management and authentication workflows across 6+ modules over UART for a 3-KM, 2-node QKD test environment. I independently defined the system architecture and encryption workflow from zero pre-existing specifications, delivering end-to-end session setup in 500ms while owning 8+ component designs across authentication, key lifecycle, and QKDN integration.

On the GenAI side, I engineered an AI red teaming playground (Project Nika) with a multi-stage defense pipeline that evaluates prompt injection, jailbreaks, encoding exploit, and authority override patterns before and after LLM invocation. I built Docex AI, a local-first multimodal knowledge platform that supports 20+ file formats with OCR and speech-to-text, using FAISS + RAG to reach sub-2s query latency, and I’ve validated my security mindset through hackathon and CTF performance.