We don't talk about AI, we run it in production
Morpheus AI serving 50+ clients, plugged.in with 2,700+ MCP server registry, TÜBİTAK-backed cognitive memory R&D. We combine AI infrastructure engineering with enterprise infrastructure experience since 2004.
Morpheus AI — Production in 50+ Clients
Morpheus is VeriTeknik's AI assistant integrated into the Ops Hub platform. Server ordering, domain management, service monitoring, invoice queries — automating infrastructure operations with 30+ AI tools.
30+ AI Tools
VPS ordering, domain management, invoice queries, service monitoring, payments, SMS 2FA — ActionContext architecture
Cognitive Memory
6 memory rings: fresh → long_term → habits → procedures → shocks → dos_and_donts. Reinforcement-based lifecycle, BIOS layer
Dual-Process Processing
Focus Agent (System 1, ~100ms): ikigAI local model. Analytics Agent (System 2): deep analysis. Kahneman dual-process theory
plugged.in — 2,700+ MCP Servers, Open Source
plugged.in is an open-source platform that manages AI agent memory, tools, and knowledge base. Actively used by 1,062+ developers.
Jungian Archetype System
"Friday afternoon deploys fail 3.4x more often"
"Docker volume permission error: chmod 755 on host dir"
"Deploy → verify → rollback sequence for your K8s pipeline"
"This namespace config silently drops health checks under load"
Our GPU, Our Model
Local inference with vLLM tensor parallelism on dual RTX 3090 (48GB VRAM). Morpheus's Focus Agent runs here — client data doesn't leave, KVKK compliant.
- Hardware: Dual RTX 3090, ASUS ROG Maximus XIII Hero Z590
- Software: vLLM, LiteLLM proxy
- Model: Gemma 4 E2B-IT (Focus Agent), Qwen3 series
- Routing: LiteLLM → vLLM (local) or Claude/GPT (external)
zvec — Sub-Millisecond Vector Search
Based on Alibaba's Proxima engine, zvec is the vector search infrastructure for plugged.in and Morpheus AI. Migrated from Milvus — superior in performance and operational simplicity.
CogMem-AI — Cognitive Memory Architecture for AI Agents
CogMem-AI is a cognitive memory architecture inspired by Kahneman's dual-process theory and McClelland's complementary learning systems theory.
Dual-Process Cognitive Memory
System 1 (~100ms) and System 2 accessing the same substrate without conflict
Reinforcement Lifecycle
Promotion chain calibration with recall × success across 6+ rings
Knowledge Engineering Inference Engine
Conflict management between LLM probabilistic output and deterministic rules
Collective Gut Agent
Cross-tenant pattern clustering under differential privacy
Prof. Dr. Turgay Çelik (University of Agder, CAIR) · Dr. Kutluhan Erol (İzmir Ekonomi Üniversitesi)
2 SCI papers · 1 patent application (TPE + EPO) · TRL 5→7
PAP — Open Source Agent Lifecycle Protocol
An open protocol managing the discovery, authorization, execution and termination lifecycle of autonomous AI agents. Apache 2.0 licensed.
Dual-profile: PAP-CP (gRPC/mTLS) + PAP-Hooks (JSON-RPC/OAuth)
AI/ML Ops Services
We bring our product-proven experience to our clients.
RAG System Setup
Smart answers from your documents — with plugged.in Knowledge Base.
- Document upload and chunking with plugged.in KB
- Semantic matching with zvec vector search
- Hybrid recall: semantic + keyword search
- Token budget management and context optimization
MCP Integration
Connect the 2,700+ MCP server ecosystem to your agents.
- MCP server discovery via plugged.in registry
- Custom tool development
- PAP Protocol compliant agent integration
- Multi-provider routing (LiteLLM)
AI Agent Development
Production-ready agent architecture with Morpheus reference.
- Proven ActionContext architecture with 30+ tools
- Cognitive memory integration (CogMem)
- Development with Claude Agent SDK
- Autonomy levels and security layers
LLM Infrastructure
Local or cloud — KVKK compliant inference infrastructure.
- ikigAI local inference (dual RTX 3090, vLLM)
- LiteLLM multi-provider routing
- Scalable inference with tensor parallelism
- KVKK compliant local deployment option
Vector Database
Sub-millisecond vector search with zvec.
- zvec (Alibaba Proxima) — production-grade performance
- Migration experience from Milvus
- pgvector — PostgreSQL integration for RAG v2
- Embedding model optimization
Our Open Source Projects
| plugged.in App | AI infrastructure platform | MIT |
| plugged.in Plugin | Claude Code plugin, memory + archetypes | MIT |
| plugged.in MCP Proxy | 2,700+ MCP server proxy | MIT |
| PAP | Agent Protocol | Apache 2.0 |
| Kintsugi | AI-native content annotation | MIT |
13 open source repos under VeriTeknik GitHub organization.
Why VeriTeknik — AI/ML Ops
Let's Design Your AI Infrastructure Together
Let's evaluate your AI/ML needs. RAG, agents, inference infrastructure — what's the priority?
- AI infrastructure needs analysis
- Current stack assessment
- Schedule a technical call
AI infrastructure needs analysis
Prefer direct contact?