Nathan El-Khoury
Honours BCS. of Computer Science; AI & ML Stream
Carleton University
Technologies: PyTorch, TensorFlow, Scikit-learn, Pandas, NumPy, ReactJS, NestJS, Next.js, Flask, .NET, Blazor, Node.js
Databases/Tools: MySQL, MongoDB, FAISS, Chroma, PostgreSQL, Git, AWS, Azure, Docker, Kubernetes, Linux, Unix
I'm a focused software engineer with a passion for AI, machine learning, and building technology that solves real-world problems. My skills have been shaped by impactful projects and internships across ML and full-stack development. Outside of tech, I'm interested in finance, reading, Muay Thai, and lifting weights.
Education
Carleton University
Honours BCS. of Computer Science; Math Minor
GPA: 3.97/4.0
Stream: AI & ML Stream
Relevant Courses: Artificial Intelligence, Machine Learning, Neural Networks, Database Management, Data Structures and Algorithms, Algorithm Analysis and Design, Operating Systems, Statistics, Linear Algebra
Experience
Machine Learning Engineer Intern
Python, FastAPI, Flask, LangGraph
- Led development of a multi-agent LLM compliance system converting natural language queries into on-demand visual reports for 10,000+ customers, cutting delivery time from weeks to under 30s
- Orchestrated a LangGraph workflow that uses OpenAI APIs for NLP, MCP data retrieval, and LLM-based visualization, auto-generating PDF reports and autonomously resolving 90% of user queries
- Built MCP servers and a multi-server router mapping intents to 4 MCP servers and 60 APIs with 95% accuracy
Machine Learning Engineer Intern
Python, PyTorch
- Built a Python synthetic data framework, boosting data variety, raising model accuracy 25% and cutting false positives 30%
- Engineered a multi-object tracking evaluation pipeline with dynamic ReID in Python, computing MOTA/HOTA and IDF1 metrics for granular model performance insight
- Developed a CNN-LSTM rock fall classifier in Pytorch, achieving 92% accuracy under varying environmental conditions
Software Engineer Intern
Python, Typescript, NextJS, NestJS
- Optimized computer vision pipelines for edge devices by introducing parallel computing, increasing FPS by 30%
- Developed a real-time video analytics interface in NestJS/Next.js with AWS S3 integration, decreasing latency by 3s and scaling concurrency 5x for ML-powered security systems
- Led development of a rockfall classification system and ETL pipeline; live conference demos secured $500k+ in contracts
Software Engineer Intern
C#, Python
- Implemented NLP-driven keyword extraction using TF-IDF and word co-occurrence, boosting enterprise processing 5×
- Architected the Datahunter Query Language using a customized command-pattern backend and integrated it into the Blazor frontend, enabling complex user queries while accelerating resolution time by 200%
- Overhauled the data catalog pipeline in Databricks, resulting in secure and automated data exports to Azure storage
Software Engineer Intern
Python, C++, Typescript
- Developed an SVM-based anomaly detection system using k-fold cross-validation, achieving 98% accuracy and reducing manual testing time by 30% in optical transceiver quality assurance
- Automated optical tranceiver report generation using LLM pipelines, cutting manual effort by 15 hours/week
- Built a ReactJS/Flask dashboard for hardware control, real-time diagnostics, and automated reports to unify workflows
Portfolio
Machine Learning UFC Predictor
Engineered features and built an ensemble ML model (XGBoost + Logistic Regression) on historical fight data, deployed via ReactJS + Flask, delivering real time predictions with 86% accuracy
Computer Vision Poker Game Theory Optimizer
Built YOLOv8+ByteTrack card tracking pipeline that streams detections into Monte Carlo simulations to compute win odds and optimal next move (fold / call / raise)
MRI Brain Tumor Detection
Developed a CNN-based tumor detection system matching DenseNet’s 86% accuracy at 1/3 the computational cost.
Health and Fitness Club Management System
Designed a web application that facilitates interaction among various user categories including members, trainers, and administrators. This platform replicates essential functions needed to manage a fitness club, like personal logging, scheduling member and trainer sessions, organizing classes, and maintaining equipment.
Workout Tracker
Created a web application that enables users to register or log in and monitor intricate workout routines, encompassing varied exercise plans, specific sets, repetitions, and weights. This application helps users precisely track their fitness journey and observe their progress over time.
Real Time Chat Application
Developed a real time chat application which allows users to create an account and chat with each other.
Algorithmic Infared Image Detection
Created a python application which takes in infared image frames and detects heat signatures.