Goal: Build an industrial fleet management platform for Autonomous Mobile Robots (AMR) in factory and warehouse environments.
Approach: Full-stack platform with FastAPI + SQLAlchemy backend, React 19 + TypeScript frontend, real-time telemetry, A* pathfinding, and Leaflet-based interactive map editing.
Result: Supports 5 robot states, 3 mission execution modes (sequential, parallel, mixed), JWT-based role access control, and a built-in robot simulator for hardware-free testing.
Goal: Create a full-stack web platform for 3D visualization, physics simulation, and hardware control of robotic arms (KUKA KR210).
Approach: React 19 + Three.js for real-time 3D rendering, PyBullet for physics simulation, Pyodide for in-browser Python execution, ROS2 WebSocket bridge, and pyFirmata for Arduino servo control.
Result: Supports drag-and-drop URDF loading, teach pendant interface, workspace reachability analysis, Monaco Editor for scripting, and session recording.
Goal: Design a real-time closed-loop PID controller for a roll-forming machine to regulate sheet metal thickness at 0.835 mm.
Approach: Modeled the full system in MATLAB/Simulink with a DC motor (2nd-order state-space), gear train, and ODE3 Runge-Kutta solver. Auto-generated C code via Simulink Embedded Coder for ST10 microcontroller deployment.
Result: Stable control with Kp=3.0, Ki=6.0, Kd=5.0, anti-windup derivative filtering (N=100), and deterministic 0.5 s interrupt-driven execution.
Goal: Develop an intelligent weekly task scheduler that optimally distributes tasks while respecting deadlines and workload constraints.
Approach: Three-stage hybrid pipeline — Fuzzy Logic scoring (urgency, priority, contribution, effort), Shannon Entropy + WASPAS MCDM weighting, and Genetic Algorithm optimization (population=50, generations=200, tournament selection).
Result: Balances soft penalties for daily overload and hard penalties for deadline violations, producing optimal 7-day schedules for custom task sets.
Goal: Build a comprehensive data-driven inventory management system to address stockouts, overstock, supplier risks, and profitability issues.
Approach: 12 integrated analysis modules covering ABC classification, Economic Order Quantity (EOQ), demand forecasting, supplier risk scoring, seasonality detection, and cross-selling analysis on 500 products and 150K+ transactions.
Result: Identifies that A-class products generate 79.84% of revenue; avg EOQ of 74 units; peak sales in August; 28.6% of inventory flagged for urgent action. Outputs Markdown, LaTeX, and PDF reports.
Goal: Create a lightweight, offline-capable Windows desktop application for writing and previewing UML diagrams using PlantUML syntax.
Approach: Built with C# + WPF on .NET 8, embedded plantuml.jar for self-contained rendering (no internet required), 16 pre-built diagram templates (Class, Sequence, Use Case, Activity), and skinparam theme customization.
Result: Supports PNG/SVG export, multilingual UI in 13 languages (including Turkish & German), runs on Windows 10/11 x64.