Menu

The Empathic
Simulation.

Client Celebrity Cruises
Partner CyberDyme (USA)
Core Tech Eye Tracking + NLP

We engineered a sentient training environment that doesn't just listen—it watches. By fusing Eye-Tracking Biometrics with Voice Emotion AI, we quantify empathy.

BIOMETRICS: ACTIVE VOICE: PHONEME ANALYSIS

7

Emotional Tones

60Hz

Gaze Tracking

1M+

Phonemes Trained

AWS

Secure Backend

Hospitality is Invisible.

Celebrity Cruises (via CyberDyme, USA) faced a unique problem: How do you train "empathy"? Standard roleplay is inconsistent, and video tutorials are passive.

They needed a way to objectively measure the intangible: Eye contact, tone of voice, and micro-expressions. Riad Saad (TopCode) was tasked with architecting a system that could digitize these human signals into actionable data.

Biometrics: Tobii / SRanipal Eye Tracking Voice AI: PyTorch Phoneme Analysis Engine: Unity (HDRP -> URP Optimized) Cloud: AWS (EC2/Fargate, DynamoDB)

The "Behavioral Engine"

We built a feedback loop that trains the subconscious.

1. Gaze Telemetry

Using headset eye-tracking, we measure "Time to Contact." Did the trainee look the guest in the eye? Did they notice the dirty glass on the table? We generate heatmaps of attention.

2. Phoneme & Tone Analysis

We trained a custom PyTorch model on 1M+ voice segments. It doesn't just check what you said, but how you said it—detecting frustration, hesitation, or warmth.

3. Adaptive Scenarios

The AI Guest reacts to your biometrics. If you avoid eye contact, the guest becomes annoyed. If you speak calmly, the guest de-escalates. It's a living simulation.

The Biometric Loop

Turning raw human signals into structured performance data.

SENSORS

Gaze + Audio Stream

ANALYSIS CORE

Emotion Classification

DASHBOARD

Behavioral Heatmaps

"This ecosystem was engineered in strategic partnership with CyberDyme (USA), with Riad Saad serving as Lead Architect for the AI & VR implementation."

Collaboration Credit
Full-Cycle Execution by TopCode Founder

Need to Train the Unspoken?

Request Architecture Review