Negar Honarvar Sedighian

Working closely with dentists to develop AI models for dental assistants, integrating clinical expertise into the model development process.

Developing phishing detection systems and contributing to a comprehensive AI assistant fine-tuned for Farsi.

Managed access to company resources using Windows Azure; identified and analyzed potential network weaknesses and breaches.

Completed a network security bootcamp covering security vulnerabilities, risk identification, incident response, and penetration testing.

Shahid Beheshti University · Tehran, Iran
Cumulative GPA: 17.49/20 (3.68/4) · Final two years: 18.32/20 (3.79/4)

Selected courses (GPA 4/4): Computer Vision, Robotics, Deep Reinforcement Learning, Machine Learning, Artificial Intelligence, Algorithms Design, Computer Simulation, Statistics and Probability, and Data Structures.

Farzanegan 1 High School (SAMPAD) · Mashhad, Iran
GPA: 19.95/20 (4/4). Ranked first in the SAMPAD high-school entrance examination in Khorasan Razavi Province (2017).

github.com/negarhonarvar/DAGMRNet

• Proposed a dynamic attentive GNN for reconstruction of highly undersampled cardiac MRI images from the CMRxRecon dataset.
• Models spatial and temporal relationships using image self-similarity; uses PromptUnet, introduced in PromptMR, for sensitivity map estimation.

github.com/negarhonarvar/StockTrading_DRL

• Designed a cascading LSTM–PPO model to capture temporal dependencies in stock data and optimize trading decisions.
• Used a yfinance-based environment with trading data from January 2009 to June 2024.

GitHub repository

• Developed an enhanced neural network to classify Farsi news articles into categories.
• Prepared the dataset with tokenization and feature extraction.

GitHub repository

• Designed a controller for autonomous flight over boxes in an urban environment, image capture, and deep-learning-based object recognition.
• Programmed the quadcopter to land beside a matching target and activate its front LEDs.

github.com/negarhonarvar/Bug-Algorithms

• Implemented Bug1, Bug2, and wall-following algorithms for GCTronic’s e-puck in Webots to navigate a maze.
• Generated a maze map using Bug2 and split-and-merge algorithms.

github.com/negarhonarvar/TronGame

• Combined a genetic algorithm with Minimax as its fitness function for a real-time two-agent game.
• Built on a Unity framework based on Chillin’s monitor games, with collision avoidance and competitive wall creation.