pro
Mahmut Düzen
Founding AI Engineer & Product Architect
New to Contra
Mahmut is ready for their next project!
Personalitic — AI-Powered Synthetic Persona & Ad Simulation Platform
Multi-tenant SaaS platform that simulates marketing campaign performance using synthetic personas and reinforcement learning. Designed for marketing teams to test campaigns before launch.
Platform capabilities:
Multi-tenant SaaS combining commercial product modules with AI orchestration
Synthetic persona simulation workflows for campaign scenario analysis
Reinforcement learning pipelines to simulate ad performance outcomes
Subscription, usage tracking, and tenant management infrastructure
Real-time analytics dashboard for campaign insights
Stack: Node.js, TypeScript, Python, PyTorch, Nuxt 3, PostgreSQL, Redis, Docker, GraphDB.
Role: Founding engineer and architect. Designed multi-tenant SaaS infrastructure, built persona simulation engine, implemented RL pipelines, developed backend services.
0
23
PlaySense — AI Sports Analytics & Player Tracking
Computer vision platform that autonomously tracks football players and generates performance analytics from amateur match footage.
Technical capabilities:
Multi-camera video stitching and synchronization
Real-time player detection and tracking with YOLOv8 + DeepSORT
Action recognition pipelines for pass, shot, and tackle detection
Player heatmap generation and distance metrics
OCR-based jersey number recognition for player identification
Solved real-world challenges: occlusion handling, jersey similarity, motion blur
Stack: Python, PyTorch, TensorFlow, OpenCV, YOLOv8, DeepSORT, FastAPI, PostgreSQL, MongoDB, Docker.
Role: Computer vision engineer. Designed video intelligence pipelines, implemented detection and tracking algorithms, developed analytics outputs.
0
27
AhiCommerce — AI-Powered Multi-Tenant E-Commerce Platform
Architected and built a comprehensive e-commerce platform integrating 30 AI/ML models for retail intelligence. Solo development from concept to production deployment.
Platform capabilities:
13 microservices handling product catalog, orders, inventory, recommendations
30 integrated AI models: product matching, price optimization, demand forecasting, customer segmentation, fraud detection
Multi-tenant architecture supporting isolated workspaces
Real-time analytics with ClickHouse for billion-row aggregations
Multi-currency, multi-language support
Stack: .Net Core, TypeScript, Python, PostgreSQL, MongoDB, ClickHouse, Redis, RabbitMq, Docker, Kubernetes, gRPC.
Role: Sole architect and engineer. Designed the entire platform, built all 13 microservices, integrated all AI models, and deployed production infrastructure.
0
35
PromptWall — Enterprise AI Security Platform
Multi-tenant AI security firewall protecting enterprise LLM usage. Built end-to-end as sole architect and engineer: 34 modules, 5 custom Python AI models (NER, Embedding, Injection Detection, Toxicity, OCR), multi-tenant control plane. Key features: Multi-LLM integration with circuit breaker pattern (OpenAI, Anthropic, Google, Ollama, vLLM), semantic DLP using pgvector + HNSW indexing, document protection with OCR-based image analysis, browser/editor/CLI/agent guard surfaces. Stack: Node.js, TypeScript, Python, PostgreSQL with pgvector, gRPC, Next.js, Docker, ClickHouse. https://promptwall.net
0
29