AI MODEL DISCLOSURE

StudLLM: Our Core AI Strategy

Student Learning Language Model - Built on Our Own Infrastructure

Kamhey is actively developing StudLLM (Student Learning Language Model), our proprietary large language model designed specifically for student relocation services. StudLLM represents our long-term AI strategy and core commitment to building AI capabilities on our own infrastructure.

This includes our own servers, AI agents, and workflow systems, giving us complete control over data processing, model training, and system architecture. StudLLM is being developed to provide superior performance for student-specific use cases while maintaining the highest standards of data privacy and security.

Our Infrastructure:

Own servers and compute infrastructure
Proprietary AI agents and workflow systems
Custom model training and fine-tuning capabilities
Complete data sovereignty and privacy control

OpenAI and other third-party models listed below are temporary solutions used during our beta and prototyping phase only. They will be phased out as StudLLM development progresses and reaches production readiness.

Technical AI Disclosure

✅ Complete transparency about the AI technologies, models, and systems powering Kamhey's relocation platform.

AI Model & Technology Disclosure

Last Updated: January 2025

This page provides detailed information about the artificial intelligence systems, models, and technologies we use to power Kamhey's student relocation services.

AI Systems in Use

1. Large Language Models (LLMs)

Interim Models (Beta/Prototyping Phase): OpenAI GPT-4, GPT-3.5-turbo (via API integration) - Temporary solutions during development
Purpose: Chat assistance, query understanding, content generation, translation
Provider: OpenAI LP (San Francisco, CA, USA)
Status: These third-party models are interim solutions used during our beta and prototyping phase. They will be replaced by StudLLM (our proprietary Student Learning Language Model) as development progresses.

2. Recommendation Algorithms

Technology: Custom machine learning models for housing recommendations
Purpose: Personalized property matching based on preferences and budget
Training: Proprietary models trained on anonymized user interaction data

3. Fraud Detection Systems

Technology: Anomaly detection and pattern recognition algorithms
Purpose: Identify fraudulent listings, fake profiles, and scam attempts
Provider: Stripe Radar (for payment fraud) + internal ML models

Data Processing & Training

Only privacy-compliant, non-sensitive information is processed.

What Data We Use

Anonymized and aggregated usage data
Navigation patterns
Interaction data

What We DON'T Use

Payment information
Identity documents
Sensitive personal data
Personal chat content for training purposes

Model Performance & Accuracy

We continuously monitor and report on AI system performance:

~92%

Chat Response Accuracy

Based on user satisfaction ratings and human review

~85%

Recommendation Match Rate

Students booking properties from AI recommendations

<5%

False Positive Rate

Fraud detection accuracy (legitimate users flagged)

Metrics (Examples): Evaluation is conducted through manual reviews, user feedback, and automated tests.

Known Limitations & Risks

Hallucinations: LLMs may occasionally generate plausible-sounding but incorrect information
Context Limitations: AI may not fully understand complex, nuanced situations requiring human judgment
Language Constraints: Performance varies across languages; English responses are most accurate
Legal/Medical Limitations: AI cannot provide legal advice, medical guidance, or official visa determinations
Real-Time Data Gaps: AI knowledge has cutoff dates and may not reflect the most recent policy changes

Third-Party AI Providers (Interim Solutions)

During our beta and prototyping phase, we use the following third-party AI services as temporary solutions. These will be phased out as StudLLM development progresses. All services comply with GDPR and international data protection standards:

OpenAI (GPT Models)

Provider of Large Language Models for chat and content generation

Data Processing AgreementGDPR CompliantNo Training on User Data

Stripe Radar

Machine learning for payment fraud detection and prevention

PCI DSS Level 1GDPR CompliantPayment Security

Kamhey does not sell personal data to these providers.

Your Rights Regarding AI

Users can:

Right to Human Review: You can request human review of any AI-generated decision affecting your account
Right to Opt-Out: You can disable AI chat features and request human-only support
Right to Explanation: You can request an explanation of how AI recommendations or decisions were made
Right to Data Exclusion: You can opt out of having your data used for AI model improvement

Exercise Your Rights

Contact: support@kamhey.com

Model Updates & Changes

We regularly update our AI systems to improve performance and safety. Significant changes are announced via:

Email notifications to registered users
In-app announcements for major model upgrades
Updates to this disclosure page (with version history)
Monthly transparency reports (available on our blog)

Current Model Version: v2.1 (January 2025)

Latest update: Enhanced multilingual support, improved fraud detection accuracy

Questions about our AI models?

For questions about our AI technologies, model performance, or disclosure information:

AI Disclosure Contact:

Email: support@kamhey.com

Last Updated: January 2025 | Kamhey Corporation, London, United Kingdom