RITIKA SHARMA View Resume ↗

Cross - camera deep search

Role
UI/UX Designer
Scope
UX/UI & Rapid Prototyping
Category
Enterprise Web App
Industry
Security & Surveillance Tech
CamPulse deep search results screen layered over the area-of-interest selection drawer

1.Product Context & Vision

CamPulse is an enterprise Video Management Software (VMS) built to manage thousands of cameras across massive facilities like airport terminals, corporate campuses, and transit hubs.

Instead of traditional security systems that passively record video, CamPulse uses an automation-first, operator-centric approach where we aim to reduce operator eye strain, eliminate complex menus, and turn raw video feeds into actionable security alerts.

The Goal: Redesign the video search experience so security teams can track individuals across multiple cameras in seconds.

2.The Problem

Security guards monitor dozens of camera feeds simultaneously. When an incident occurs (like a theft or unauthorized entry), finding where a suspect came from or where they went requires hours of manual timeline scrubbing across separate camera screens.

  • High Risk of Missing Threats

    Guards cannot watch every screen at once.

  • Slow Incident Response

    Manual searching takes up to 45 minutes per incident.

  • Clunky Legacy Interfaces

    Traditional tools force users to jump between separate viewing and searching pages.

3.Research & Insights

To build this 0-to-1 feature, we interviewed control room operators and security heads to understand how investigations happen under stress:

Visual Context > Text Queries

Operators identify suspects by visual markers ("man in red jacket") rather than camera IDs or exact timestamps.

Speed is Everything

The transition from identifying a suspect on a live feed to searching for their previous/subsequent locations must happen in under 3 clicks.

False Positive Fatigue

Search results must display visual match confidence scores so operators can quickly ignore weak matches without re-watching full clips.

4.Rapid Prototyping: From "Vibe Coding" to Figma

To design our multi-camera search layout quickly without wasting days drawing static screens, we "vibe coded" early interactive prototypes using Google AI Studio. It allowed us to test real interactive behaviors in a live browser environment.

Once the interaction logic was proven with stakeholders, we designed the validated flows into Figma, aligning every component with our CamPulse design system.

5.The Final Experience & Step-by-Step Flow

1

Spotting a Suspect on Live View

When a guard spots suspicious activity on a live video tile, selecting that tile brings up a contextual control panel on the right with a prominent, primary "Deep Search" CTA button.

UX Impact:

Zero menu diving. The action lives directly next to camera controls.

2

Area of Interest Selection

Clicking Deep Search opens a focused drawer with an adjustable, draggable bounding box around the suspect or object.

UX Impact:

Instead of forcing guards to type text descriptions (e.g., "man in brown jacket"), guards can simply crop the person. The backend AI engine instantly extracts the visual features.

3

Multi-Camera Results & Evidence Collection

The guard is brought to the Deep Search results screen, pre-filled with extracted visual tags (Male, White Shirt, Age 34) and a synchronized grid showing all cameras where that suspect appeared.

UX Impact:

Each video tile displays an explicit Match Confidence Score (e.g., 94% match) and a primary "Download Evidence" CTA, letting guards package video clips for police in seconds.

The three screens of the deep search flow: live view, area of interest selection, and multi-camera results
Live view with the Deep Search action in the camera control panel
Area of Interest Selection drawer with a draggable bounding box
Deep Search results grid with match confidence scores and download evidence

6.Impact & Outcomes

Clear improvements in investigation speed, system ease of use, and guard focus after launching the new flow.

90% Faster Investigations

Cut suspect tracking time across 100+ cameras from hours of manual video scrubbing to under 30 seconds.

Near-Zero Training Required

Replaced complicated search filters with an intuitive "Crop & Search" visual interaction that any guard can use on day one.

Reduced Cognitive Load

Guards maintain spatial and visual context throughout the entire process without jumping through disjointed software tabs.

7.Lessons Learned

Key insights gained from rapid AI prototyping and real-world testing with control room operators.

AI Confidence Matters

Showing clear match percentages (like 94% match) gives operators immediate confidence to trust AI results without re-watching full clips.

AI Prototyping Accelerates UX

Using AI Studio to prototype interactive concepts early saved massive design iteration time before finalizing Figma design system components.