01
Private knowledge
Retrieve the right internal context without treating every interaction as disposable chat history.
AI-first product startup • Working MVP
Governed enterprise AI for decisions that can be traced and controlled.
Mert Bileydi MBAI Automation is building an enterprise reasoning platform that connects private knowledge, model reasoning, human approval and auditable actions.
Validated today
32B AWQ reasoning core served with vLLM
- Self-hosted 32B model inference
- Secure gateway and controlled action boundary
- Initial identity and prompt-injection test flows
The 72B QLoRA model is the next development milestone. It is not presented as a completed production model.
Working MVPValidated 32B model core
Product companyNot an agency or reseller
Startup ecosystemAlanya TEKMER
CredentialTeknogirişim Rozeti
The product
From private company knowledge to governed AI action.
The MBAI AGI-Oriented Enterprise Reasoning Platform is a digital-native software product. K.U.B.R.A. Genesis V14 is its internal reasoning architecture and technical codename.
02
Governed reasoning
Combine model output with policy, identity, evaluation and explicit human approval gates.
03
Auditable workflows
Connect reasoning to tools and business processes while preserving reviewable action records.
Who it is for
Knowledge-intensive teams that need private, controlled and reviewable AI assistance.
What it replaces
Disconnected chat tools, manual context transfer and unsafe unapproved automation.
What makes it AI-first
The reasoning model, retrieval layer and approval architecture are the product’s foundation—not an add-on.
Working MVP
A clear separation between what works now and what comes next.
The current product state is presented precisely so reviewers, partners and future customers can evaluate it without inflated claims.
Validated now
Working
- Self-hosted 32B AWQ model core
- vLLM inference service
- Secure application gateway
- Initial identity test flow
- Initial prompt-injection test flow
Active development
Improving
- Private retrieval quality and chunk selection
- Persistent project memory
- Approval policies and audit trails
- Enterprise workflow integration
Next milestone
Planned
- 72B QLoRA model development and evaluation
- High-memory GPU experiments on cloud infrastructure
- Multimodal evaluation
- Customer pilot environment
Operational evidence
Lina Kaya case study
An AI operations case study used to test campaign planning, analytics, audience growth, link conversion and outreach workflows. It demonstrates the broader controlled-workflow direction; it is not presented as the enterprise product itself.
28.3K+Ad impressions tracked
+433Follower growth measured
42+Outreach targets
Business model
A software company with repeatable product revenue.
MBAI Automation is not a development shop, consultancy, cloud reseller or outsourced agency. Revenue is planned around access to its own enterprise AI product.
Enterprise SaaS
Recurring workspace subscriptions based on team size, governance requirements and product capabilities.
API access
Usage-based access for approved reasoning, retrieval and workflow calls integrated into customer systems.
Paid pilots and private instances
Time-boxed enterprise pilots and private deployment options for organizations with stronger data-control needs.
Current commercial stage:
pre-revenue product development and pilot preparation. The company plans to seek venture funding after the next model, product and customer-validation milestones.
Founder and company
Mert Bileydi
Founder & Chief Architect
Mert leads product architecture, model evaluation, infrastructure planning and commercialization for MBAI Automation. The company was founded in May 2026 as a sole proprietorship in Alanya, Antalya, Türkiye.
Product roadmap
Measured progress from model core to enterprise pilots.
-
1
32B prototype
Self-hosted model serving, secure gateway and initial security test flows.
-
2
Reasoning platform
Retrieval quality, memory, approval policies and auditable workflow components.
-
3
72B QLoRA milestone
Cloud GPU training experiments, evaluation and deployment optimization.
-
4
Enterprise pilots
Private customer environments, measurable pilot outcomes and repeatable onboarding.
Why Google Cloud
Infrastructure for product development—not the product being resold.
Planned uses include high-memory model evaluation, QLoRA experiments, retrieval infrastructure, secure demo deployment, observability and customer pilot environments.
- Vertex AI / Compute Engine GPU evaluation
- Cloud Storage for controlled model artifacts
- Cloud Run for application services and demos
- Monitoring and data analysis for evaluations
Contact
Discuss the product, a pilot or a startup program review.
Mert Bileydi — Founder & Chief Architect