My Experiences

Thesis: Architecting a Deep Learning-Based Digital Twin and Reinforcement Learning Control Pipeline for Wind Turbines.

🔗 View the Repository: github.com/itreza7/master-thesis

I designed and deployed an end-to-end, data-driven optimization framework for cyber-physical systems, graduating Rank 1 in my cohort. The core objective was to maintain a stable 5MW power output under highly stochastic environmental conditions without relying on computationally expensive physics simulations. To achieve this, I first developed a temporal forecasting module utilizing a BiLSTM network with a 2-hour sliding window to predict upcoming wind dynamics, validating its accuracy with rigorous RMSE and $R^2$ metrics.

To solve the computational bottlenecks of agent training, I constructed a dynamic surrogate environment—a localized Digital Twin. I used K-means++ clustering to divide the operational data into distinct regimes, training specialized Multi-Layer Perceptrons (MLPs) for each cluster to accurately simulate the physical turbine's power generation. Finally, I trained a Rainbow DQN reinforcement learning agent. To ensure the physical safety of the turbine's mechanical actuators, the agent utilized a continuous-to-discrete incremental action space ($\Delta \theta$) and a squared-error reward function to optimize blade pitch. The entire architecture was built with rigorous, industry-standard software engineering practices, bridging theoretical control methodologies with scalable, deployment-ready code.

As a Technical Team Lead, I directed multidisciplinary teams across DevOps, frontend, and backend development to align technical roadmaps with core business objectives. I balanced strategic leadership—defining project goals and optimizing development workflows—with high-level execution, such as architecting a Next.js multi-tenant platform integrated with a complex ecosystem of 66 Python-based microservices. Leveraging AWS and FastAPI, I ensured seamless cross-team collaboration and the delivery of scalable, high-quality infrastructure.

As a Senior Backend Developer, I architected and deployed a highly scalable ecosystem of 66 Python-based microservices using FastAPI, designed to support over 1 million daily users. I engineered the internal service communication and unified them behind a secure Gateway API using AWS Gateway API, while developing a comprehensive dashboard featuring robust RBAC (Role-Based Access Control). To ensure high-performance data management across diverse geographic locations, I implemented a sophisticated polyglot persistence layer leveraging PostgreSQL, MongoDB, Redis, and OpenSearch.

In my role as a Multimedia Teaching Assistant, I mentored students in applying Deep Learning and Computer Vision principles using the Python ecosystem. I provided technical guidance on architecting CNNs and DNNs with PyTorch, while integrating OpenCV for image processing and TorchAudio for signal analysis. Additionally, I facilitated the use of Scikit-Learn for traditional machine learning workflows and Gradio to build interactive interfaces for complex neural network models.

As a Laboratory Instructor at Amirkabir University of Technology, I led the Logic Circuit Laboratory, focusing on the practical application of hardware description and digital design. I provided technical instruction in VHDL and the programming of FPGA boards, guiding students through the design, simulation, and physical implementation of complex logic systems.

As a freelance Back End Developer for nearly six years, I specialized in the design and implementation of web applications, modules, and scripts using PHP and Laravel. I managed the full project lifecycle for diverse clients, frequently leading small development teams to ensure technical alignment and high-quality delivery. My work focused on building robust backend systems, including custom WordPress solutions and efficient web scraping tools, while maintaining seamless collaboration across remote environments.

As a Teaching Assistant for Digital Systems I at the University of Guilan, I supported Dr. Shahbaz Reyhani in delivering core curriculum to undergraduate students. I facilitated learning through weekly tutorial sessions and the development of custom educational slides and problem sets. Beyond instruction, I provided technical mentorship for final projects using Proteus simulation software and managed the end-to-end grading process for assignments and exams.