Sina Heydari

AI Researcher

I am a computer scientist and AI researcher working on machine learning, deep learning, and representation learning.

My research focuses on developing robust and structured machine learning methods for learning meaningful representations from complex data.

I am particularly interested in combining neural architectures with structured representations to improve the robustness, interpretability, and generalization of AI systems.

University email · Gmail · ORCID · Google Scholar · LinkedIn · GitHub

Education

M.Sc. in Computer Engineering, Artificial Intelligence and Robotics
Institute for Advanced Studies in Basic Sciences (IASBS)

B.Sc. in Computer Engineering, Software Engineering
Azad University of Zanjan

Research Interests

Machine Learning · Deep Learning · Natural Language Processing · Representation Learning · Graph Neural Networks · Large Language Models

Research Experience

Research Assistant, Natural Language Processing Lab

February 2026 – July 2026
Department of Computer Science and Information Technology
Institute for Advanced Studies in Basic Sciences (IASBS)
Supervisor: Dr. Majid Ramezani

  • Designed a multi-level hypergraph framework for modeling personality traits as overlapping, weakly supervised categories.
  • Benchmarked the framework against deep learning baselines and analyzed performance under label noise and class overlap.
  • Prepared and submitted a manuscript to a peer-reviewed journal.
  • Provided research guidance and methodological support to junior students joining the laboratory.

Paper Presentation

2025
RSAttAE: An Information-Aware Attention-Based Autoencoder Recommender System
11th International Conference on Web Research (ICWR 2025)

Paper · Presentation Certificate

CS & IT Department Representative and Presenter

2025
26th National Exhibition of Research and Technology Achievements

  • Developed and presented an LLM-based personality prediction system from text.
  • Built an object detection application using YOLOv8.

CS & IT Department Representative and Presenter

2024
25th National Exhibition of Research and Technology Achievements

  • Designed and implemented an LLM-based Persian poetry contest application, addressing language-specific challenges in NLP.

Publications

HyperPersona

HyperPersona: A Multi-Level Hypergraph Framework for Text-Based Automatic Personality Prediction
Preprint

A hypergraph-based framework for modeling hierarchical textual representations for automatic personality prediction.

arXiv · Code

CDAE

CDAE: Enhancing Perturbation Robustness in Pretrained Language Models with Contrastive Denoising
Submitted to the 16th International Conference on Computer and Knowledge Engineering (ICCKE 2026)

A contrastive denoising framework for learning sentence representations that remain robust under semantically preserving textual perturbations.

arXiv · Code

RSAttAE

RSAttAE: An Information-Aware Attention-Based Autoencoder Recommender System
Published at the 11th International Conference on Web Research (ICWR 2025)

An information-aware recommender system for rating prediction, designed to capture the relative importance of user and item features for recommendations.

Paper · Code

Teaching Experience

Teaching Assistant

March 2024 – June 2024
Department of Computer Science and Information Technology, IASBS
Course: Large Language Models
Instructor: Dr. Majid Ramezani

  • Designed and supervised assignments on autoregressive versus autoencoding models, GPU architectures for LLM training, and the mathematical formulation of attention mechanisms.
  • Designed and supervised projects on training an encoder-only model, building a chatbot with a free API, evaluating generation metrics including BLEU, ROUGE, and CIDEr, and zero-shot classification using CLIP.
  • Helped students debug code for course projects.
  • Held office hours and review sessions and graded assignments and examinations.
  • Curated relevant research papers and assisted students in preparing presentations.

Teaching Assistant

September 2024 – December 2024
Department of Computer Science and Information Technology, IASBS
Course: Text Mining and Web Mining
Instructor: Dr. Majid Ramezani

  • Designed and supervised assignments on building a simple search engine, investigating similarity measures in graph space, and Byte Pair Encoding.
  • Designed and supervised projects on building a DFA-based tokenizer, training classifiers including Decision Tree, Naive Bayes, SVM, KNN, and ensemble methods, and implementing clustering algorithms including K-Means and cluster representation on the 20 Newsgroups dataset.
  • Helped students debug code for course projects.
  • Held office hours and review sessions and graded assignments and examinations.
  • Curated relevant research papers and assisted students in preparing presentations.

Academic Projects

Finding the Shortest Path Using A-star Algorithm

GitHub Repository · Winter 2024
Course: Multi-Agent Systems (MAS)

Implemented the A-star (A*) algorithm from scratch to determine optimal routes between nodes in a grid-based environment, enabling autonomous agents to navigate dynamically while avoiding collisions and optimizing route planning.

Exploratory Data Analysis (EDA) and Classification

GitHub Repository · Winter 2024
Course: Machine Learning (ML)

Developed machine learning models to predict whether an online advertisement click leads to a purchase, distinguishing purchase-inducing clicks from non-converting interactions to support profit maximization.

Classification Models

GitHub Repository · Winter 2024
Course: Machine Learning (ML)

Implemented weighted logistic regression using Newton-Raphson optimization and analyzed the effect of the locality parameter τ. Compared Naive Bayes and SVM classifiers on the MNIST dataset using One-vs-All and One-vs-One strategies.

Principal Component Analysis (PCA)

GitHub Repository · Winter 2024
Course: Computational Data Mining

Implemented PCA from scratch in Python for dimensionality reduction and visualization, including Scree plots, Score plots, and Loading plots for analyzing feature contributions.

Transformer and iTransformer

GitHub Repository · Spring 2024
Course: Deep Learning (DL)

Implemented and compared Transformer and iTransformer architectures from scratch for sequence modeling tasks, demonstrating the efficiency and performance characteristics of the iTransformer architecture.

Ontology-Based Knowledge Representation

GitHub Repository · Spring 2024
Course: Natural Language Processing (NLP)

Developed an ontology-driven graph framework for semantic knowledge representation, enabling structured modeling, reasoning, and retrieval through custom ontology construction and semantic querying.

Sentiment Analysis Using Naive Bayes Binary Classifier

GitHub Repository · Spring 2024
Course: Natural Language Processing (NLP)

Implemented an NLP pipeline for classifying IMDB reviews using a Naive Bayes classifier developed from scratch and a scikit-learn baseline. Evaluated performance using accuracy, F1 score, recall, confusion matrices, and ROC curves.

Kernel Principal Component Analysis (KPCA)

GitHub Repository · Spring 2024
Course: Pattern Recognition

Implemented Kernel PCA in MATLAB to demonstrate nonlinear dimensionality reduction through kernel-based implicit mapping into feature space.

Self-Organizing Map for Clustering and Visualization of Iris Data

GitHub Repository · Fall 2024
Course: Artificial Neural Networks (ANN)

Implemented a Self-Organizing Map from scratch in MATLAB for unsupervised clustering and visualization of the Iris dataset, with quantitative evaluation using quantization, distortion, and topographic error metrics.

Skills

Programming

Proficient: Python, PyTorch, PyTorch Geometric, NetworkX, Hugging Face, Scikit-learn, NumPy, Pandas, spaCy, NLTK, Matplotlib, LaTeX, MATLAB, NetLogo, C#

Familiar: SQL, Java, R, Julia

Development Tools

Python: VS Code, Spyder, JupyterLab, Jupyter Notebook, Google Colab

LaTeX: TeXstudio, TeXmaker, Overleaf

Operating Systems and Software

Operating Systems: Linux (Ubuntu), Windows

Software: LM Studio, Cursor

Languages

Farsi: Native

English: Advanced (B2–C1)

Certifications