Mahsa(Zahra) Anvarian

Computer Vision Researcher

May 2024 - Present

University of Calgary, Canada

Vision Research Lab
Supervisor: Dr. Farhad Maleki

• Demonstrated that random tile-based splitting understated DEM RMSE by 10% through spatial data leakage, establishing region-based evaluation as a more reliable protocol for assessing real-world generalizability.
• Designed frequency-aware loss functions combining wavelet, terrain, and geometric cues (slope, aspect, TRI, TPI), improving fine-detail reconstruction by 6% on TRI RMSE.
• Collected, built, and cleaned spatially disjoint, terrain-diverse datasets of ~26K paired 120 m/30 m SRTM DEM tiles spanning mountainous, moderate, and low-relief regions, using Python, Rasterio, and NumPy.
• Adapted, trained, and benchmarked SOTA super-resolution architectures in PyTorch under controlled experimental protocols.
• Developed the full research pipeline end-to-end (geospatial acquisition, preprocessing, patch generation, training, evaluation, and reproducible experiment management) and created a Polyscope-based Python package for interactive 3D inspection of DEM tiles.

Natural Language Processing Researcher

Sep 2020 - June 2021

Iran University of Science and Technology, Iran

NLP Research Lab
Supervisor: Dr. Sauleh Eetemadi

• Studied the literature of datasets and models designed for psychological traits prediction.
• Constructed the first dataset based on MBTI model of personality traits for Persian language.
• Implemented multiple Neural Network models to predict users' personality traits using their tweets.
• Experienced coding NLP models using fastText, doc2Vec, BERT, CNN, LSTM, etc.

Computer Vision Researcher

May 2019 - Sep 2020

Iran University of Science and Technology, Iran

Multi Agent Systems (MAS) Lab
Supervisor: Dr. Nasser Mozayani

• Took Computer Vision Coursera course for learning more about CV applications.
• Implemented a Neural Style Transfer using VGG-19, which is a pre-trained model, to injected the style to the input image.
• Designed a Car Detection model using YOLO algorithm to recognize cars in images with the bounding boxes.

Machine Learning Engineer

Sep 2020 - Oct 2022

Tehran, Iran

Dadmatech Company

• Built the first large-scale Persian dataset for MBTI personality-trait prediction, collecting and processing 1.55M+ tweets from 938 users using Python and the Twitter API, later expanded to 6.1M+ tweets from 3,876 users.
• Developed data pipelines for user validation, tweet collection, Persian-language filtering, anonymization, preprocessing, and class balancing using Pandas, NumPy, NLTK, and Hazm.
• Fine-tuned ParsBERT using HuggingFace Transformers and developed a combined fastText-embedding, BiLSTM, and attention-based model across four MBTI classification tasks in TensorFlow/Keras, benchmarked against BERT, Doc2Vec, CNN, and LSTM baselines.
• Achieved 66.71% average accuracy across the four MBTI tasks, outperforming the majority-class baseline by approximately 13 points on P/J and 7 points on T/F.
• Applied LIME to interpret model predictions and surface the influential Persian words and user-level linguistic patterns driving each trait, turning a black-box classifier into an auditable one.

Front-End Developer

Jul 2019 - Dec 2019

Remote (Ontario, Canada)

Edgecom Energy Company

• Built two interactive dashboards for real-time electricity demand and peak-load monitoring in JavaScript, jQuery, and HTML/CSS, visualizing usage at yearly, monthly, and daily granularity via live charts, historical peak tables, and a probabilistic “chance of peak” gauge.
• Integrated a weather-forecasting API to correlate temperature and humidex against demand, and built comparative views (actual vs. IESO-projected demand, multi-day similarity curves) supporting near-term peak-load estimation.

Teaching Assistant at University of Calgary

May 2024 - Apr 2026

Calgary, Canada

Excellent TA Award, 2025/2026

• Taught tutorials and labs for four data science and AI courses over four terms, reaching 15+ students from first-year undergraduates to graduate level; distilled deep learning, model evaluation, and distributed data processing into material each cohort could act on.
• Mentored 10+ student teams through semester-long course projects end-to-end — scoping, big-data processing pipelines, and cloud deployment on AWS — running weekly checkpoints and unblocking design, debugging, and scalability issues from proposal through final delivery.
• Ran office hours supporting Python, TensorFlow, Pandas, Tableau, and Power BI workflows, diagnosing broken pipelines and overfitting issues under time pressure, and partnered with faculty to keep tutorial content aligned with lectures.

CPSC 433: Artificial Intelligence
Instructor: Dr. Jonathan Hudson

DATA 608: Developing Big Data Applications
Instructor: Dr. Tyler Bonnell

DATA 201: Thinking with Data
Instructor: Dr. Andy Asare

DATA 601: Working with Data and Visualization
Instructor: Dr. Fateme Rajabiyazdi

Teaching Assistant at Iran University of Science and Technology

Feb 2018 - Jun 2022

Tehran, Iran

Natural Language Processing
Instructor: Dr. Behrooz Minaei

Computational Intelligence
Instructor: Dr. Nasser Mozayani

Deep Learning
Instructor: Dr. Mohammad Reza Mohammadi

Microprocessor and assembly language
Instructor: Dr. Amir Mohammad Monazzah

Embedded System and IoT
Instructor: Dr. Amir Mohammad Monazzah

Database Design
Instructor: Dr. Eisa Zarepour

Mentoring

May 2019 - Aug 2020

Software Engineering
Instructor: Dr. Mehrdad Ashtiani

Feb 2018 - Jun 2018

Computer Systems Analysis and Design
Instructor: Dr. Mehrdad Ashtiani

EPA Conference 2026

Methodological Factors Affecting the Generalizability of Deep Learning-Based Digital Elevation Model Super-Resolution [PDF]
Z. Anvarian, F. Maleki

• Studied how dataset construction and evaluation protocol affect the real-world generalizability of deep-learning-based DEM super-resolution models.

ICBEB Conference 2026

Beyond Mean Pooling: Statistical Meta-Aggregation for Subject-Level EEG Classification in Parkinson's Disease [PDF]
P. Taghipour*, A. Khodadadi*, F. Rezaei†, Z. Anvarian†, R. C. Sotero, S. Madadi

• Proposed a statistical meta-aggregation approach for subject-level EEG classification in Parkinson's disease, improving over standard mean pooling.

WiNLP Conference 2022

MBTI Personality Prediction Approach on Persian Twitter [PDF]
S. Fatehi*, Z. Anvarian*, Y. Madani, M. J. Mehditabar, S. Eetemadi

• Collected a dataset that contained Persian tweets as data and MBTI personality traits as the label.
• Implemented a Bidirectional RNN+LSTM(64) model for predicting users' MBTI personality traits.

WeCNLP Conference 2021

ParsTSet: A Persian Dataset for Personality Detection on Twitter [PDF]
M. M. Abdollahpour*, Z. Anvarian*, S. Fatehi, S. Eetemadi

• Proposed a novel dataset along with a baseline model for predicting MBTI personality traits.
• Working on the full paper, so it is an ongoing research project.

Computer Vision

Car Plate Detection GitHub

• Trained multiple deep CNN models to classify images that contain car plates.
• Ran a 15-configuration ablation over architecture (Xception, ResNet50, MobileNet), optimizer, and augmentation for three-class license-plate classification on an imbalanced 2,727-image dataset; HSV lighting augmentation reached 99.64% accuracy (0.99 macro-F1), while synthesized minority-class data degraded every architecture.
• Implemented relatively complex deep models using Keras.
• The best project of the class both in terms of execution time and F1-score.

Deep Neural Network GitHub

• Trained ResNet50 model to classify the Stanford Car dataset images using Keras, and Augmented data with data generator to reduce overfitting.

Convolutional Neural Network GitHub

• Designed a CNN model, which includes an Inception module with dimension reduction, using Keras to classify the Fashion MNIST dataset images.

Image Classification via shape, texture, and color GitHub

• Coded the HOG and LBP, which are image feature extractors, using OpenCV to classify the MNIST dataset images by SVM classifier.

Computational Intelligence

Inverted Pendulum GitHub

• Solved Inverted Pendulum using Fuzzy Logics (also using RL in Gym env).

Image Classification GitHub

• Designed a Multi-Layer Perceptron (MLP) model to classify the Hoda dataset images, which like MNIST dataset but in Persian, using Numpy and Keras.

HopField Network GitHub

• Implemented a noise-robust model using Hopfield Network for image detection.

Radial Basis Function GitHub

• Coded the function approximation using RBF (Radial Basis Function) and MLP.

Self-Organizing Feature Map GitHub

• Trained a Kohonen’s Self-Organizing Feature Map (SOFM), which can map a dataset of 3-Dimensional data into a 2-Dimensional space.

Artificial Intelligence

News Classification GitHub

• Trained two models, Naïve Bayse and MLP model, that classify news documents into two classes: politics news and sport one.

Reinforcement Learning GitHub

• Implemented Reinforcement Learning in games like WaterWorld or PixelCopter

AI Pacman game GitHub

• Solved Pacman practical Projects of Berkeley University in the most of AI outlines such as Search Problems, Informed Search, CSP, Adversarial Search, Markov Decision Process, etc.

Signal Processing

Digital Radio GitHub

• Designed a digital radio, which can detect radio channels and play them using Signal Processing.

Dual-Tone Multi-Frequency GitHub

• Implemented a Dual-Tone Multi-Frequency (DTMF) signaling that each of the 12 keys on the phone sends a specific signal when clicked.

Yes-No Signal Detection GitHub

• Implemented the Yes-No signal detection that get the voice of Yes or No and detect its signal.

deeplearning.io

Oct 2020

Certificate

Convolutional Neural Networks in TensorFlow

Oct 2020

Certificate

Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning

Oct 2020

Certificate

Natural Language Processing in TensorFlow

Sep 2020

Certificate

Deep Learning Specialization

Sep 2020

Certificate

Sequence Model

Aug 2020

Certificate

Convolutional Neural Networks

Aug 2020

Certificate

Structuring Machine Learning Projects

Aug 2020

Certificate

Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization

Jun 2020

Certificate

Neural Networks and Deep Learning

Coursera

Mar 2022

Certificate

Deep Learning with PyTorch : Neural Style Transfer

Feb 2022

Certificate

Deep Learning with PyTorch : Generative Adversarial Network

Feb 2022

Certificate

Deep Learning with PyTorch : Image Segmentation

Sep 2020

Certificate

Computer Vision - Image Basics with OpenCV and Python

Elsevier

Apr 2022

Certificate

Becoming a Peer Reviewer

Jun 2022

Certificate

Certified Peer Reviewer Course