We design edge vision pipelines that detect, track, and classify objects or gestures in real time—on resource-constrained devices:

• Vision AI on MCUs and FPGAs: Running optimized CNNs using platforms like Lattice CrossLinkU, ARM Cortex-M, or ESP32-S3
  
  
• Gesture Recognition: Dynamic hand gesture classification for control interfaces or wearables
  
  
• Object Detection & Tracking: Real-time inference for people, faces, signs, or custom-trained classes
  
  
• Sensor Fusion: Combine IMU, ToF, or proximity sensors to enhance robustness
  
  
• Power-Efficient Design: Always-on vision with sub-10mW standby modes Lattice SensAI, custom quantized models, image preprocessing in FPGA fabric

tinyVision.ai delivers robust, always-on voice interfaces for wearables and smart devices:

• Wake Word Detection: Low-power keyword spotting with fast wake-up
  
• Command Recognition: On-device inference of voice commands (custom or general-purpose)
  
• Voice Triggered UI: For hands-free control in AR/VR, smart glasses, or IoT
  
  

We build camera pipelines that combine sensor control, AI processing, and data streaming—all in embedded form factors:

• MIPI to USB/UVC Camera Modules
  
  
• Onboard AI Processing: Face recognition, pose estimation, presence detection
  
  
• Custom ISP & Preprocessing: Debayering, resizing, RGB/YUV conversion in FPGA or MCU
  
  
• UVC Camera Firmware: Plug-and-play integration with host systems
  
  
• Stereo & Depth Capable: Modular camera designs for depth-aware applications

• Deep expertise in embedded systems design
  
  
• Full-stack capabilities: sensor → firmware → AI model → application
  
  
• Proven performance on MCUs, FPGAs, and SoMs
  
  
• Active in tinyML, edge vision, and voice-AI communities