← Back to Projects
Product · MVP · Web Platform · 2026

VisionSense

Defined and launched VisionSense from scratch, a sophisticated vision AI system spanning drone defense, invisible environmental monitoring, and industrial inspection.

Achieved Product-Market Fit by securing strategic integrations with Samsung and Quickset, leveraging Atomrock's Edge AI foundations to solve complex B2B challenges and expanding the enterprise pipeline by 65%.

Role

Product Builder
Definition · Design · MVP Development

Company

Atomrock

Type

Commercial B2B Product
Cloud & On-Prem · MVP Deployed

Stack

Claude Code · YOLO
AI Detection · Model Training

Watch full drone demo

Atomrock's Edge AI foundations shipped with general-purpose detection capabilities (Motion AI and Face AI) running across diverse edge devices. VisionSense is the sophisticated, domain-specific product I built on top of that foundation to tackle complex B2B challenges basic detection alone couldn't solve.

I owned the full product definition end-to-end. I scoped what VisionSense should be, chose which three application categories to ship in the MVP, defined how data flows between edge and cloud, and shaped how model configuration, training, and inference surface to non-technical operators.

Three application categories I defined for the MVP

  1. Drone intrusion detection for aerial defense.
  2. Particle, gas, and thermal visualization for invisible environmental monitoring.
  3. Industrial inspection for counting and defect detection on production lines.
Drone intrusion detection for aerial defense — live tracking with thermal overlay and security-zone alerts
Drone Defense · Live Demo MVP
Particle, gas, and thermal visualization for invisible environmental monitoring — thermal zone configuration
Invisible Environmental Monitoring · AI Configuration (Thermal Mode) MVP
Industrial inspection for counting and defect detection on production lines — PCB defect detection with object and sub-feature AI models
Industrial Inspection · AI Model Configuration (Object + Defect) MVP

What I defined

  1. Product scope: moving beyond general Motion and Face AI into a domain-specific, full-cycle vision system.
  2. Data pipeline: how training data is collected, how model outputs are transmitted in real time, and how events are logged for later review.
  3. Display layer: real-time monitoring canvas, event review, snapshot playback, and model configuration panels.
  4. Operator workflow: a path for non-technical users to train and configure AI behavior for their own use cases.
  5. Multi-domain architecture: one system that scales across drone defense, environmental monitoring, and industrial inspection.

Product, Design & Engineering

Wanling Yu

Company

Atomrock

Year

2026

Platform

Cloud & On-Prem Web
(MVP Deployed)

Next Project
AtomView