class AIResearchEngineer:
    """
    Senior Computer Engineering student specializing in the convergence of 
    geometric deep learning, generative design, and autonomous systems.
    Career trajectory: Full-Stack Development → AI/ML Research Engineering
    """
    
    def __init__(self):
        self.identity = {
            "name": "Sirac Gezgin",
            "location": "Bursa, Turkey",
            "institution": "Bursa Technical University",
            "graduation": 2026,
            "current_role": "AI R&D Intern @ Martur Fompak International",
            "research_focus": [
                "3D Generative Design & FEA",
                "Neural Implicit Representations",
                "Computer Vision for Industrial QA",
                "Edge AI & LLM Deployment"
            ]
        }
        
    def current_research(self) -> dict:
        """Active research at Martur Fompak International AI R&D Lab"""
        return {
            "project": "Neural Implicit Representations for Finite Element Analysis",
            "reference": "arXiv:2110.10863",
            "methodology": [
                "Deep Learning architectures from scratch (PyTorch)",
                "PointCloud Processing & Topology Optimization",
                "Voronoi Diagrams for structural analysis",
                "Open3D framework for 3D data manipulation"
            ],
            "industrial_application": {
                "domain": "Automotive Quality Control",
                "system": "Light Guide Vision System",
                "technology": "YOLOv8-Pose for operator tracking",
                "scale": "24 quality checkpoints, real-time inference"
            }
        }
    
    def philosophy(self) -> str:
        return """
        Bridging theoretical AI research with production-grade engineering.
        Transforming complex mathematical models into deployed, real-world systems.
        From academic papers (arXiv) to industrial applications (automotive, robotics).
        """

# Initialize
researcher = AIResearchEngineer()
print(researcher.philosophy())

Career Narrative: Deliberately pivoting from full-stack web development (React, Node.js, Angular) to specialized AI research. Leveraging strong software engineering fundamentals to build robust, production-ready machine learning systems. Not just a model trainer—a systems architect who understands the full pipeline from data to deployment.

October 2025 - Present | Bursa, Turkey

Research Objective: Investigate neural implicit representations as computationally efficient alternatives to traditional Finite Element Analysis in structural optimization.

Technical Implementation:

• Literature Foundation: Deep dive into arXiv:2110.10863 on neural implicit fields
• Framework Mastery: Implementing core PyTorch components via d2l.ai curriculum
  • Custom backpropagation engines
  • Neural network architectures from scratch
  • Gradient descent optimization algorithms
• 3D Data Pipeline:
  • PointCloud processing with Open3D
  • Topology optimization workflows
  • Voronoi diagram generation for structural analysis
  • Integration with FEA simulation data

Key Technologies: PyTorch Open3D NumPy PointNet Neural Radiance Fields

Industrial Challenge: Real-time verification of 24+ quality checkpoints on automotive seat assembly lines with sub-second latency requirements.

Solution Architecture:

graph LR
    A[Industrial Camera Array] -->|1080p @ 30fps| B(Image Preprocessing)
    B -->|OpenCV Pipeline| C{YOLOv8-Pose Model}
    C -->|Keypoint Detection| D[Operator Action Classifier]
    D -->|Procedural Compliance| E[Quality Verification Engine]
    E -->|Pass/Fail Signal| F[Assembly Line Control]
    
    style C fill:#00FFFF,stroke:#333,stroke-width:3px,color:#000
    style E fill:#76B900,stroke:#333,stroke-width:2px,color:#fff

Technical Contributions:

• Pose estimation model research & evaluation (YOLOv8 architecture)
• Real-time inference optimization for industrial hardware
• Camera interfacing and image processing pipelines (OpenCV)
• Operator movement analysis algorithms for 24 quality checks

Impact: Automated quality assurance replacing manual inspection, reducing human error and increasing throughput.

Continuous Learning: Monitoring cutting-edge AI developments

• Google ADK (Agent Development Kit) for voice agents
• Hugging Face Transformers & Agents certification courses
• LLM fine-tuning techniques for domain-specific applications

August 2025 - October 2025 | Özveri R&D Center

Role Context: Agile team collaboration on internal web applications

Technical Stack:

• Angular (TypeScript, SCSS)
• Modular UI component architecture
• Responsive design implementation
• UI/UX standards adherence

Key Outcome: Experience in production software development lifecycle, reinforcing engineering discipline applied now to AI systems.

graph LR
    A[Mic] -->|Audio| B(Whisper)
    B -->|Text| C{TinyLlama}
    C -->|Intent| D[Home API]
    C -->|Response| E(TTS)
    E --> F[Speaker]
    
    style C fill:#EE4C2C,color:white

# INT4 quantization pipeline
model = load_quantized_model(
    "TinyLlama-1.1B-Chat",
    quantization="int4",  # 3GB → 750MB
    device="cuda:0"
)

Environment: Gazebo (Full Physics)
Framework: ROS/ROS2 Node Architecture
Sensors: LiDAR, IMU, Cameras
Testing: Hardware-in-the-Loop (HITL)

graph TD
    A[Traffic Cameras] --> B[YOLO Detection]
    B --> C[Vehicle Tracking]
    C --> D[LSTM Model]
    D --> E[Predictions]
    E --> F[Traffic Control]

Frameworks:
  Primary: PyTorch (Production-grade)
  Secondary: TensorFlow, Scikit-learn
  
Specializations:
  - Neural Architecture Design
  - Custom Loss Functions
  - Model Optimization & Quantization
  - Transfer Learning Strategies
  
Advanced Topics:
  - Graph Neural Networks (GNNs)
  - Neural Radiance Fields (NeRF)
  - Generative Adversarial Networks
  - Large Language Models (LLM)
  
Tools & Libraries:
  - Hugging Face Transformers
  - Weights & Biases (Experiment Tracking)
  - ONNX & TensorRT (Optimization)
  - Ray Tune (Hyperparameter Tuning)

Core Libraries:
  - OpenCV (Industrial Applications)
  - Ultralytics YOLOv8 (Object Detection)
  - Open3D (3D Vision)
  
Techniques Mastered:
  - Object Detection & Tracking
  - Pose Estimation (Human/Object)
  - Image Segmentation
  - PointCloud Processing
  - 3D Reconstruction
  
Domain Applications:
  - Industrial Quality Control
  - Autonomous Navigation
  - Medical Imaging
  - Robotics Perception

Edge AI:
  Hardware: NVIDIA Jetson Nano
  Optimization: CUDA, TensorRT, ONNX
  Deployment: Docker, Model Serving
  
Programming Languages:
  Expert: Python (AI/ML focus)
  Proficient: C/C++ (Performance-critical)
  Working: SQL (Data pipelines)
  
Development Tools:
  - Git & GitHub (Version Control)
  - Ubuntu/Linux (Primary OS)
  - Jupyter/VS Code (Development)
  - Docker (Containerization)
  
Robotics & Simulation:
  - ROS/ROS2 (Robot Operating System)
  - Gazebo (Physics Simulation)
  - PX4 Autopilot (Drone Software)

Data Manipulation:
  - Pandas (DataFrames)
  - NumPy (Numerical Computing)
  - Matplotlib/Seaborn (Visualization)
  
NLP Libraries:
  - Transformers (State-of-the-art models)
  - Zemberek (Turkish NLP)
  - spaCy (Text Processing)
  
Specialized Skills:
  - Time-Series Forecasting
  - Topology Optimization
  - Statistical Analysis

Contribution Activity Graph

Language Distribution

Connect for Research Collaborations & AI Projects

Professional Profile

siracgezgin@gmail.com

Open Source Work

Educational Videos

"Bridging theoretical AI research with production-grade engineering."