Data Scientist

Summary

Data scientist bridging radiation therapy expertise with advanced deep learning, specializing in the design, training, and optimization of neural network architectures (U-Net, CNNs, RNNs, Transformers). Skilled in model evaluation, feature engineering, knowledge distillation, and model compression. Proficient in Python, PyTorch, Scikit-learn, and pandas, with a strong ability to build reproducible pipelines and deliver actionable insights from data. Currently completing a Master’s in Computer Science (Data Mining), focusing on advanced model architectures and performance optimization.

Core Competencies

Deep Learning (CNNs, RNNs, U-Net, Transformers); Model Optimization; Knowledge Distillation; Python (PyTorch, Scikit-learn); Statistical Modeling; Predictive Analytics; Feature Engineering; Data Mining; ML Pipelines; Data Visualization

Education

MASTER’S DEGREE IN COMPUTER SCIENCE – DATA MINING |

TARBIAT MODARES UNIVERSITY – TEHRAN 2024 – (EXPECTED GRADUATION 2026)

In third semester studying Computer Science (Data Mining) Relevant Coursework:

• Deep Learning: MLPs, CNNs, RNNs, Transformers; implementation in PyTorch for image classification, segmentation, and sequence tasks
• Machine Learning: Supervised and unsupervised learning algorithms, model evaluation, regularization techniques
• Statistical Analysis: Hypothesis testing, probability distributions, statistical modeling
• Data Mining: Pattern recognition, data preprocessing, feature engineering

BACHELOR’S DEGREE IN RADIATION THERAPY |

BABOL UNIVERSITY OF MEDICAL SCIENCES - BABOL 2020 – 2024 GPA: 17.4

• Participated in a cohort study on cancer patients, managing and entering psychological data using SPSS software.
• Exposure to healthcare systems and data handling, particularly in cancer treatment research.
• Worked in clinical environment as a radiation therapy specialist for two semesters.

Technical Skills

• Languages:
  • Python(Proficient)
  • SQL(Basic)
  • Git(Basic)
• Machine Learning and Deep Learning:
  • Supervised Learning: Logistic Regression, SVM, Decision Trees, Random Forest, Gradient Boosting
  • Unsupervised Learning: K-Means, Spectral Clustering, UMAP
  • Deep Learning: CNNs, RNNs, Transformers, U-Net architectures
  • Model Evaluation & Optimization: Hyperparameter tuning, cross-validation, regularization
  • Advanced Techniques: Knowledge Distillation, Composite Loss Functions, Triplet Loss
• Data Analysis:
  • Data Manipulation: Pandas, NumPy
  • Data Visualization: Matplotlib, Seaborn
  • Statistical Analysis: Hypothesis testing, correlation analysis
• Frameworks and Libraries:
  • ML/DL Frameworks: PyTorch, Scikit-Learn, Hugging Face
  • Data Tools: SQLite, Git
  • Computer Vision: OpenCV
  • Web Applications: Streamlit

Certificates

Intermediate Machine Learning - 1403
Data Manipulation with Pandas - 1402
SAT 1490/1600 - 2022

Language Skills

• English - Advanced (Toefl 111/120)
• Persian – Native

Socials

Github \ Linkedin

Projects

U-net Reconstruction with Knowledge Distillation

• Developed a sophisticated U-Net architecture for simultaneous image reconstruction and segmentation using knowledge distillation
• Implemented composite loss function (MSE + feature distillation + fuzzy normalized cut loss) to improve performance
• Utilized pre-trained VGG11 as teacher model to guide learning process
• Achieved strong performance metrics: PSNR of 23.23 dB and SSIM of 0.859 on CIFAR-10 dataset, performing well considering the low epoch runtime of the model
• Technologies: PyTorch, U-Net, Knowledge Distillation, VGG11, Composite Loss Functions

Deep Learning Foundations

• Implemented SVM/SVR as neural networks, and MLPs and CNNs from scratch on MNIST dataset.
• Developed and tested multiple MLP configurations on MNIST dataset, analyzing effects of hidden layers, neurons, and activation functions
• Built 2-layer CNN with pointwise convolution for efficient MNIST classification achieving 82% accuracy in 25 epochs
• Successfully implemented SVM and SVR as neural networks with 88% accuracy and 0.2233 MSE respectively
• Technologies: PyTorch, CNN, MLP, SVM/SVR, Neural Networks

Classification with Confusion Analysis using UMAP

• Built CNN classifiers with cross-entropy and triplet loss, achieving 84% accuracy on MNIST in 30 epochs.
• Performed comprehensive confusion matrix analysis to identify frequently confused digit pairs (classes 2 & 5)
• Implemented UMAP dimensionality reduction for visualizing feature embeddings and analyzing class separation
• Technologies: PyTorch, CNN, Triplet Loss, UMAP, Confusion Matrix Analysis

Superstore Sales Data Analysis

• Conducted full data cleaning, preprocessing, EDA, and clustering on 10k+ sales records.
• Used K-Means to segment customers into 3 key behavioral clusters, supporting targeted marketing.
• Developed data visualizations and insights to support targeted marketing strategies
• Technologies: Python, Pandas, Matplotlib, Scikit-Learn, K-means Clustering

Model Separability and Noisy Data Classification

• Compared multiple models (Logistic Regression, Linear Regression(Lasso, Ridge), SVM, Bayes, Naïve Bayes, Perceptron) across linearly separable and noisy datasets.
• Visualized decision boundaries and evaluated algorithm robustness to noise.
• Provided insights into model selection based on data characteristics and noise levels
• Demonstrated systematic approach to algorithm evaluation and understanding of theoretical foundations
• Technologies: Python, Scikit-Learn, Classification Algorithms, Data Visualization