Trust Infrastructure · B2B AI

Truth
travels fast

FYNX AI detects hallucinations, misinformation, and fraudulent AI-generated feedback before it reaches your users, your ratings, or your regulator.

FAKE · 0.94
AUTHENTIC · 0.91
AI-GENERATED · 0.87
FYNX AI

Reveal the true rating behind manipulated feedback

Most rating systems are broken. They aggregate fake alongside real. FYNX AI separates the signal from the synthetic, at scale, in real time.

Hallucination Detection
Identify factual drift, unsupported claims, and AI confabulation at inference time.
Fake Review Scoring
Score every review for authenticity, flag AI-generated content, surface fraud patterns.
EU Compliance
Audit-ready outputs for Omnibus Directive (EU 2019/2161). Verified authentic consumer feedback.
Trust Signal Extraction
Surface the truthful rating hidden within manipulated feedback datasets.
Review Analysis · Live⚠ FLAGGED
"Absolutely perfect experience. Staff were incredible, rooms immaculate. Would highly recommend to everyone without hesitation!"
AI-Generated
0.88
Sentiment Manip.
0.76
Vague Phrasing
0.82
Authenticity
0.12
Adjusted True Rating
3.7 / 5.0
▼ 1.3 below displayed average
Fraud Removed
34%
of review volume

Fake feedback is a business risk, not a moderation problem

AI-generated and fraudulent reviews erode trust, skew decisions, and create regulatory exposure.

01
Regulatory pressure is accelerating
EU Omnibus Directive (2019/2161) mandates verified authenticity. Fines up to 4% of annual revenue or €2M, whichever is higher. Enforced since May 28, 2022.
02
AI makes fraud effortless
AI-generated reviews on TripAdvisor grew 137% between 2019 and 2024, from 4.49% to 10.7% of all reviews. Legacy moderation cannot keep up.
03
Trust compounds
Platforms with trusted ratings outperform on retention, conversion, and partner confidence. Booking.com alone hosts 240+ million verified reviews.
0%
of websites violating EU fake review law
Source: EC Sweep, Jan 2022
0%
of revenue or €2M: EU Omnibus maximum fine
Source: Directive (EU) 2019/2161
0%
growth in AI-generated TripAdvisor reviews (2019-2024)
Source: Originality.ai, Dec 2025
0
verified reviews on Booking.com
Source: Emily Dean Statistics, 2024

Built from inside AI evaluation

Our founders spent years inside the pipelines that trained and evaluated the models everyone now worries about.

Domain Expertise
RLHF & Hallucination Detection
Direct experience using RLHF to identify hallucinations in STEM model outputs, the exact threat we detect for clients.
Evaluation Infrastructure
Truthfulness at Scale
Rating model outputs, designing adversarial prompts, building scoring criteria at Outlier.
Technical Depth
Full-Stack AI Engineering
From 42 Heilbronn's rigorous curriculum to production deployment, we ship software that works.

Built on Google Cloud

Cloud-native on GCP: Vertex AI for training, BigQuery for analytics, Cloud Run for auto-scaling API. EU region (europe-west3) for GDPR compliance.

🧠
Model Training & Inference
Vertex AI + Cloud GPUs
Fine-tuning detection models on A100/T4 GPUs. Sub-200ms inference via Vertex AI Endpoints.
45%
📊
Data Pipeline
BigQuery + Pub/Sub
Real-time streaming ingestion and pattern detection across millions of reviews.
25%
API Infrastructure
Cloud Run + Firestore
Zero-to-scale API serving. Real-time trust scores. Auto-scaling with demand.
30%
◈ Real-Time Review Processing Pipeline
PIPELINE ACTIVE AVG LATENCY: 142ms
INGESTION
INPUT
📥
Review Submitted
OTA Platform / API Webhook / REST Endpoint
~2K req/min
MESSAGE QUEUE
GCP
📡
Cloud Pub/Sub
Real-time message queue & event ingestion
Buffered99.9% SLA
GOOGLE CLOUD
STREAM PROCESSING
GCP
⚙️
Cloud Dataflow
Stream processing · Feature extraction · Text normalization
Apache BeamAuto-scale
GOOGLE CLOUD
PARALLEL PROCESSING
VERTEX AI
🧠
Vertex AI Models
Hallucination detection · Fake review classifier · Coordination analysis
A100 GPUs<80ms
VERTEX AI · A100 GPUs
⟷ SYNC
GCP
📊
BigQuery
Analytics warehouse · Cross-platform pattern matching · Historical analysis
Petabyte-scaleReal-time
GOOGLE CLOUD
RESULTS MERGED
API LAYER
SERVING
Cloud Run API
Trust scores · Verdicts · Compliance reports · <200ms latency
Auto-scaling<200ms E2EREST + gRPC
GOOGLE CLOUD · AUTO-SCALING
VERDICT ROUTING
BLOCKED
🚫
Fake → Blocked
Fraud score > 0.7 · Auto-quarantined
Flagged for review
OR
PUBLISHED
Real → Published
Compliance trail included · EU audit-ready
Verified authentic
<200ms
Target Latency
99.9%
GCP Pub/Sub SLA
10.7%
AI Reviews on TripAdvisor
EU-West3
GDPR Region
Projected GCP Credit Allocation
Vertex AI / GPU Training
45%
BigQuery & Dataflow
25%
Cloud Run & GKE
20%
Storage & Monitoring
10%

See FYNX AI in action

Paste any review and watch our model score it in real time.

REVIEW INPUT
PROCESSING REVIEW THROUGH FYNX AI PIPELINE...
AI-GENERATED
-
AUTHENTICITY
-
VERDICT
-
// API Response

        

Where we are

Campus Founders
Resident startup at Campus Founders, Heilbronn, Germany
Pain Confirmed
Stakeholder interviews completed. Problem validated.
Pre-MVP
Architecture designed. Building toward first prototype.
2025
Founded in Heilbronn, Germany
✓ Completed
Campus Founders residency (workspace + programs)
Campus Corporate Challenge completed by CEO
AI START program completed by CEO at Campus Founders
42 Heilbronn graduation + AI specialization track
2+ years AI evaluation at Outlier (RLHF, hallucination detection)
Stakeholder interviews conducted, pain confirmed
Competitive analysis: no direct competitor in EU travel review space
→ In Progress & Next
MVP development: building first prototype
Applying for free GCP cloud credits (Google for Startups)
OTA outreach: identifying pilot partners
First pilot with OTA partner
Process 100K reviews in pilot with <200ms latency

Three co-founders who built, trained, and marketed AI.
Now they audit it.

Djallal-Eddine Mekki
Co-Founder · AI Evaluation & Vision
Engineer, developer, and AI trainer. Politecnico di Torino and 42 Heilbronn graduate. Extensive AI evaluation experience at Outlier applying RLHF to detect hallucinations in STEM outputs. Completed AI START and Campus Corporate Challenge programs at Campus Founders. Leads FYNX's evaluation methodology and product vision.
42 Heilbronn Politecnico di Torino RLHF · Outlier Hallucination Detection Campus Founders
Abdelilah Noury
Co-Founder · Engineering & Architecture
Experienced full-stack developer and 42 Heilbronn specialization student. Former AI trainer at Outlier with hands-on experience at Khyari Organization and Assala-Isskane. Designs the systems that score, classify, and surface authentic signals at production scale.
42 Heilbronn Full-Stack AI Trainer · Outlier System Architecture
Ahmed Amine Mekki
Co-Founder · Marketing & Business Strategy
Pharmacist and biotechnologist turned business strategist. Master's in Project Management & Entrepreneurship and PGS in Digital Marketing from EHEC Algiers. Consults healthcare ecosystem stakeholders and mentors medical science students launching startups. Drives FYNX's go-to-market strategy and growth.
Digital Marketing · EHEC Algiers Project Management Healthcare Consulting Startup Mentorship
"We didn't pivot into AI trust.
We came from inside it."

Djallal and Abdel met at 42 Heilbronn, a peer-driven engineering school where you build everything from scratch. Their work brought together everything they'd learned evaluating AI outputs: identifying when models hallucinate, when feedback is fabricated, and when a score cannot be trusted. Campus Founders gave them the structure to turn that insight into a company.

42 Heilbronn Campus Founders Outlier Founded 2025 Heilbronn · Germany

The trust layer
for online travel

Every booking, every rating, every recommendation is only as valuable as the authenticity behind it. We're building the infrastructure that guarantees that authenticity.

Authentic Signal Layer
Persistent trust score per review, per property, per platform, updated continuously.
Compliance by Default
Audit trails, detection logs, and verified authenticity reports for Omnibus Directive (EU 2019/2161).
The Real Rating
The score your property actually deserves, stripped of manipulation and synthetic inflation.

"FYNX AI is building the trust layer for OTA."

Want to know
what's real?

Subscribe to get updates on FYNX AI: product launches, pilot results, and insights on AI-generated fraud in online reviews.

No spam · Unsubscribe anytime · Updates only when it matters