Hi, My name is Mahdi

AWS Certified Cloud Practitioner
Apr 2024

Consultant - ML Engineer @ Deloitte
Mar 2025 - Present

Deloitte is a consulting firm providing AI solutions, along with audit, tax, and risk advisory services to clients across industries.

- Collaborated with cross-functional teams—including data engineers, analysts, and business stakeholders—to translate complex requirements into deployable machine learning (ML) solutions.
- Developed and maintained production-quality code using Python, PySpark, and SQL, supporting scalable data pipelines and machine learning models.
- Extracted and preprocessed data from both structured and unstructured sources using appropriate data preparation techniques.
- Trained and refined machine learning models to solve targeted business problems and improve decision-making.
- Deployed cloud-based ML pipelines following MLOps best practices to ensure scalability and production readiness.
- Monitored model performance through structured evaluation pipelines and presented insights in a business-friendly manner for stakeholders.

SKILLS: Python, PySpark, SQL, Azure, Databricks, MLflow, Delta Lake, Docker, Scikit-learn, PyTorch

AI/ML Developer @ Memorial University of Newfoundland
Sep 2022 - May 2025

- Explored the possibility of controlling a wheelchair through “singing imagination” by designing a Brain-Computer Interface (BCI) experiment to collect Electroencephalography (EEG) time-series dataset from 15 participants
- Achieved 85% accuracy in classifying EEG signals using various deep learning and machine learning models (CNNs, RNNs, SVM, XGBoost) and provided insights into model performance through statistical analysis
- Enhanced data quality and the labeling system by building a data preprocessing pipeline using MATLAB & Python
- Boosted classification accuracy by 8% using efficient feature extraction & optimization methods (CSP, mRMR)
- Improved data analysis speed by 50% via building an automated pipeline on Compute Canada Cloud (Code)

SKILLS: Python, R, NumPy, Pandas, Scikit-learn, PyTorch, TensorFlow, Matplotlib, Seaborn, EDA, AWS

Software Developer @ Toranj Innovation Technologies
May 2021 - Jan 2022

Toranj is a health tech company concentrated on creating software solutions for the healthcare industry.

- Elevated total revenue by 10% by building BI dashboards integrated with ETL processes using Power BI and Tableau
- Increased the customer base by 20 % by expanding system scalability through developing fault-tolerant Java microservices
- Doubled the speed of User Acceptance Testing process by implementing RESTful APIs integrated with FitNesse server
- Eliminated the need for direct SQL queries by utilizing Object-Relational Mapping (ORM) systems like JPA
- Contributed to software development in a CI/CD environment by utilizing Git for version control

SKILLS: Java, Database (PostgreSQL, MongoDB), Power BI, Tableau, ETL, Docker, Microservice Architecture

Brain-Controlled Wheelchair Using Machine Learning and Singing Imagery

In this project, we explored whether it’s possible to control a wheelchair by simply imagining singing a song in one’s head. To achieve this, we designed a Brain-Computer Interface (BCI) experiment to collect Electroencephalography (EEG) time-series data from 15 healthy participants. After data collection, I enhanced the data quality and labeling system by developing a preprocessing pipeline with MATLAB and Python. Using various deep learning models (CNNs, RNNs) and classic machine learning techniques (SVM, XGBoost), we achieved 90% accuracy in classifying the EEG signals. By applying advanced feature extraction and optimization methods (CSP, mRMR), we boosted classification accuracy by 8%. Additionally, I built an automated pipeline on Compute Canada Cloud, improving data analysis speed by 50%. We also provided statistical insights by performing hypothesis testing and analysis using R.

Medical FAQ Bot

This project develops a Medical FAQ Bot that answers medical questions using OpenAI's GPT-3.5 API and the Telegram Bot API. The bot is Dockerized and hosted on an AWS EC2 instance and operates in prototype phase, utilizing FastAPI for request processing, NLTK for user input preprocessing, and LangChain for managing conversational flow. It filters queries through a custom medical terms dataset to ensure relevance.

BI Dashboard

This project provides detailed business intelligence insights into the sales of a branch shop operating across multiple markets. The insights are derived from a SQL-based sales database containing customer information, transaction records, market names, and product details. The analysis and visualization were conducted using Tableau and Power BI with ETL integration for sales analysis, allowing for an in-depth exploration of key sales metrics and trends.

Web Scraper

This is a Python script implemented to leverage Selenium to automate the meticulous task of entering grades, feedback, and extracting class lists within the D2L platform at Memorial University. Designed specifically for Teaching Assistants (TAs) and instructors.

Key Features:
Automated Entry: Simplify your workload by effortlessly inputting grades and feedback into assignment sections and the grading area.
Class list Extraction: Streamline administrative duties with easy class list extraction.

Mini-Marketplace

Mini-Marketplace is a mini online shop in Monolithic and Microservice architecture in MVC architectural pattern, developed with the help of Spring Boot, JPA, Spring Security, and JWT.
The client can sign in, choose a product, sell his products, like other products, edit his profile, sort the products and use the OTP service.
The project has 2 panels, one for the website's admin and another for the client.
The picture is a part of the OTP service code.

IoT Hub

The project aimed to make a gateway to facilitate the commanding process in IoT without losing the connection.
One side of the gateway is connected to the network by MQTT, and the other side is connected to several devices by BLE. A client can send the command by phone, and the server will get it and pass it to the gateway through MQTT. Finally, the gateway passes the command to the connected devices by BLE. The software language of the project is C, and the hardware is NRF ARM micros(NRF51822). Right now the firmware is available on 1 to 1 connection.
This is the video for testing the MQTT side of the project, and here, it is not yet connected to the BLE.

Adaptive Harmonic IIR Filter for Frequency Estimation and Tracking

This project implements an adaptive harmonic Infinite Impulse Response (IIR) notch filter for accurate frequency estimation and tracking within a harmonic frequency environment.
The code replicates concepts from the paper 'Novel Adaptive IIR Filter for Frequency Estimation and Tracking' by Li Tan and Jean Jiang.

The primary goal of this project is to create an efficient and adaptable filter that accurately estimates and tracks frequencies in complex harmonic environments.
The implemented filter, based on the concepts presented in the paper, provides robust signal processing capabilities.

The link to the paper: Novel Adaptive IIR Filter for Frequency Estimation and Tracking

Computer Vision: Fundamental Image Filters

This code is implemented to build following filters:
1- Nearest neighbour up/down sampler
2- Convolution filter
3- High pass filter
4- Box filter
5- Bilateral filter