AmirHossein Naghshzan

Amir H. Naghshzan

PhD in Software Engineering

ML Engineer / Scientist at Replayz

Montreal, Canada

amir-h-naghshzan


Skills

Machine Learning, NLP, LLM, Deep Learning, Data Mining

Python

100%

Java

100%

Kotlin

80%

C

60%

Languages

English

Persian

French



About me

I'm Amir Hossein, a driven software engineer with a passion for ML, deep learning, and data mining. With over ten years of experience in the tech industry, I have a diverse background as an Android developer, team leader, and technical manager. Coding is my hobby, but creating impactful software solutions is my true passion.

As a Ph.D. holder in Software Engineering, I bring a wealth of knowledge and expertise to the table. My passion for AI and machine learning drives me to stay at the forefront of these cutting-edge domains, and I am constantly seeking out new challenges and opportunities to grow my skills.

Education

École de Technologie Supérieure (UQAM)
2020 - 2023

PhD of Software Engineering (GPA: 4.3/4.3)


Tarbiat Modares University
2016 - 2019

Master of Technology Management (GPA: 3.9/4)


Yazd University
2011 - 2016

Bachelor of Software Engineering


Work Experience

ML Engineer / Scientist at Replayz
May 2023 - Current

Replayz IQ is a predictive SaaS platform that uses Data Science to analyze Call Scoring Results to predict what skills matter the most for companies when it comes to increasing avg deal size, speeding up sales cycle and increasing average order size.


Software Engineer at Savoir-faire Linux
Nov 2019 - Feb 2023

Started in 1999 in Québec, Savoir-faire Linux is specialized in open-source software and digital electronics. Savoir-faire Linux is one of the largest open-source companies in Canada. I Worked on the Jami application. Jami is an instant peer-to-peer messenger for Linux, Microsoft Windows, OS X, iOS, and Android.


Head of Mobile Team at Hafhashtad
Aug 2018 - Oct 2019

Tehran Internet, also known as *780# is the largest private telecommunication company in Iran, providing diverse and up-to-date mobile payment services including charging and purchasing internet packages.


Senior Android Developer at Faraz Pardazan
Sep 2016 - Jul 2018

Faraz Pardazan is a computer software company that produces web and mobile applications related to payment and e-banking systems.


Android Developer at Parsijoo
Nov 2013 - May 2016

Parsijoo is an independent knowledge-based internet company based in Yazd, operating as a Farsi search engine. Parsijoo is Iran’s second most visited search engine after Google.


Projects

  • Twitter Like Prediction
    Hackathon (Kaggle Dataset)
    The goal was to predict the number of likes for a tweet. The final architecture was a complex combination of a BERT model trained on the tweet text and a CNN model trained on the numerical features of tweets.


  • Abstractive Summarization of Source Code
    ETS University
    We generated abstractive summaries using LLMs such as BERT, BART, and GPT for Android APIs discussed on StackOverflow and evaluated by ROUGE and BLEU.


  • Summary Generation for Code Blocks Using Their Surrounding Context
    ETS University
    We used TextRank as an unsupervised learning algorithm to generate extractive summaries for code blocks based on their surrounding natural languages such as descriptions, discussions and comments.


  • Multi-label Classification of Image Data
    Kaggle Competition - Ranked 2nd / 300 with %98 accuracy
    The competition’s goal was to develop a model to detect hand-written sequences of numbers using CNN models and data augmentation.


  • Analyzing COVID-19 Search Trends and Hospitalization on Google dataset
    Mcgill University - Mila Institution
    We were interested in analyzing and predicting the rate of COVID-19 hospitalization based on the symptoms using Google Covid-19 Open Data utilizing PCA, K-means, KNN, and decision trees.


Publications

2023

Enhancing API Documentation through BERTopic Modeling and Summarization (PDF)

AmirHossein Naghshzan, Sylvie Ratte


Improving Code Example Recommendations on Informal Documentation Using BERT and Query-Aware LSH: A Comparative Study (PDF)

Sajjad Rahmani, AmirHossein Naghshzan, Latifa Guerrouj


Leveraging Advanced Data Mining Algorithms to Recommend Source Code Changes (PDF)

AmirHossein Naghshzan, Saeed Khalilazar, Pierre Poilane, Olga Baysal, Latifa Guerrouj, Foutse Khomh


2022

Towards Code Summarization of APIs Using NLP Techniques (PDF)

AmirHossein Naghshzan


2021

Leveraging Unsupervised Learning to Summarize APIs Discussed in Stack Overflow (PDF) (Video)

AmirHossein Naghshzan, Latifa Guerrouj, Olga Baysal

2021 IEEE 21st International Working Conference on Source Code Analysis and Manipulation (SCAM)