Projects

AI-native content intelligence platform. Aggregates signals across RSS, APIs, and social feeds, distils them through LLM pipelines, and maps each insight to personal goals — turning information noise into daily actionable clarity.

Voice interface layer for AI-assisted development. Lets you talk to your editor — dictate intent, navigate context, and trigger agent actions hands-free. #23 on Product Hunt.

Obsidian plugin for semantic graph navigation. Renders force-directed clusters of vault notes, surfaces hidden conceptual connections, and supports spatial knowledge exploration with interactive filtering.

On-device accessibility app running Gemma 3n 4B via Flutter and MediaPipe. Real-time scene understanding, object detection, and depth estimation — no cloud dependency. Panasonic capstone.

Real-time inference pipeline deployed on Panasonic i-PRO cameras. INT8-quantized ONNX models run anomaly detection at sub-100ms latency. Frame-skip motion gating reduces average compute by ~70%. No cloud dependency.