Embedded & Full-Stack Systems Engineer — NasonNation Robotics
Feb 2015 — Present
• NasonNation Command - Distributed Edge/GPU Computer Vision & Biometric Security Platform:
◦ Hybrid Edge-to-GPU Inference Architecture: Engineered a containerized multi-stream RTSP/WebRTC (MediaMTX, FastAPI, React, PostgreSQL) surveillance engine running continuous YOLOv8s ONNX object detection, YuNet 5-point facial landmark alignment, and SFace 128-D biometric vector matching on an edge Linux node with asynchronous Qwen2.5-VL vision-language verification offloaded over LAN to an NVIDIA RTX 3060 Ti.
◦ Kinematic Tracking & False-Alarm Suppression: Designed a 2-pass kinematic displacement tracker with bounding-box IoU stationary footprint locking and 5-anchor polygon ray-casting, eliminating stationary vehicle/object false positives and cross-track teleportation.
◦ Zero-PII Threat Intelligence & Biometric Pipeline: Built an automated background ingestion daemon integrating the federal FBI Wanted REST API and state fugitive feeds, computing normalized 128-D facial embeddings purely in RAM (zero disk image retention) for sub-millisecond cosine-similarity watchlist matching alongside auto-clustered visitor dossiers.
• AeroTherm UAV (Thermal & LiDAR Edge Pipeline): Designed and fabricated a custom 7-inch autonomous UAV payload integrating a FLIR Lepton 3.5 and LiDAR arrays. Engineered the Python backend for dynamic, real-time thermal targeting running on a headless Raspberry Pi edge pipeline communicating over MAVLink.
• Aegis Control Sentry Turret: Architected an AI-powered physical sentry turret utilizing Roboflow and TFLite for local edge perception, real-time targeting, and mechanical tracking. Developed a Flask web interface for live telemetry monitoring.
• Autonomous UAV Fleet: Engineered custom multi-rotor UAVs utilizing Python-driven Finite State Machines (FSMs) and MAVLink telemetry for autonomous, GPS-denied indoor navigation.
• Exes Distributed Exoskeleton: Built a voice-activated prototype utilizing local edge-compute to drive 550KG servos, integrating Picovoice for offline acoustic processing, alongside a distributed MQTT telemetry network and isolated power architecture.