At work, I lead global voice-to-print infrastructure and smart device inventory replenishment at Amazon scale, alongside building Alexa device firmware.
Outside of work, I build autonomous AI applications, Model Context Protocol (MCP) infrastructure, local computer-vision servers, and zero-dependency libraries.
Led engineering for automated re-ordering and smart inventory management. Enabled connected appliances (water filters, printers, consumer dispensers) to track customer consumption in real time and autonomously reorder supplies before they run out.
Led the initiative enabling Alexa to discover, connect to, and print on virtually any printer in the world via voice. Architected the cloud document rendering pipeline, protocol negotiation, and secure device delivery at global scale.
Engineered embedded device firmware, communication protocols, and cloud-to-device synchronization systems across the Alexa hardware ecosystem and connected smart home platforms.
End-to-end applications across autonomous LLM pipelines, local computer vision, and embedded network security:
Autonomous real-time news intelligence dashboard. Fetches 15+ feeds every hour, uses Google Gemini 2.5 (with automatic Flash ↔ Flash Lite failover) to cluster stories and score sentiment, and renders an interactive D3.js treemap on Cloudflare Pages.
Privacy-first Computer Vision MCP Server for Claude. Runs YOLOv8 (ONNX) and OpenCV locally to compute sharpness maps, exposure metrics, and focus accuracy across local RAW photo libraries—paired with photo-quality-analyzer (★19).
Autonomous, always-on Home Network Sentinel and Canary Honeypot running on a sub-$4 ESP8266 microcontroller. Detects lateral network scanning, DNS spoofing, and WAN jitter with a zero-heap PROGMEM web console.
On @prasad-a-abhishek, I build tiny, single-purpose packages with zero runtime dependencies (standard library only, 39+ published on PyPI), benchmarked head-to-head against mainstream libraries: