# Navilan Santhanakrishnan > Student researcher at UC Santa Cruz (B.S. Computer Engineering, expected 2028) working on the security of AI agents. AI security research in Alvaro Cárdenas's lab, building environments that generate attacks on AI agents; industry and startup experience running evaluations for AI agents and bringing them into enterprise workflows. Joining GE Aerospace as an AI Engineer Intern in summer 2027. Website: https://navilan.dev (experience: https://navilan.dev/experience/, projects: https://navilan.dev/projects/) Email: navilan.santhanakrishnan@gmail.com GitHub: https://github.com/navilansanthanakrishnan LinkedIn: https://www.linkedin.com/in/navilansanthanakrishnan ## Experience - GE Aerospace, AI Engineer Intern (Summer 2027) - Cárdenas Lab, UC Santa Cruz, Undergraduate Researcher (2026 to now): security of AI agents; building environments that generate attacks on agents to label data for detection systems. - Greenfield Labs, Machine Learning Engineer (2026): evaluation harness for AI agents; LoRA fine-tuning of Qwen3.6-27B raised tool-call accuracy from 68% to 91%; deployed with vLLM for an enterprise client. - Going Merry, Software Engineer Intern (2026): shipped Truth24 (https://truth24.com), a news briefing product, on an ingestion pipeline for 60K+ articles. ## Projects - [qwen36-metal-research](https://github.com/navilansanthanakrishnan/qwen36-metal-research): doubled Qwen3.6-27B decode speed on an Apple M5 Pro (14.6 to 29.4 tok/s) with Metal GPU kernels and speculative decoding in llama.cpp. - [ds4-ssd](https://github.com/navilansanthanakrishnan/ds4-ssd): runs DeepSeek V4 Flash on Apple silicon by streaming mixture-of-experts weights from SSD. - [wiretrace](https://github.com/navilansanthanakrishnan/wiretrace): reverse-engineers any app's HTTP traffic into a typed OpenAPI spec and a runnable MCP server. - [exe-metal](https://github.com/navilansanthanakrishnan/exe-metal): runs Windows games on Apple silicon with Wine and D3DMetal or DXMT. ## Skills Python, C++, C, Rust, TypeScript; PyTorch, vLLM, MLX, CUDA, Metal; Docker, Kubernetes, AWS, Modal.