DEV
DALSANIA
Software Engineer • AI/ML • Montreal, QC
CS student at Concordia University building full stack applications, AI/ML systems, and production-grade infrastructure. Backend APIs, system design, and deployment end to end.
▶ ABOUT ◀
I am a Computer Science student at Concordia University who enjoys building systems from the ground up. I like working on backend-heavy problems, API design, database modeling, and deploying applications that actually run in production.
Outside of web development, I work on AI and ML projects: graph neural networks, deep learning for medical imaging, and NLP systems. I prefer structured problem solving and thinking through edge cases over just focusing on UI.
A lot of my learning comes from building projects end to end, designing the backend, connecting the frontend, and handling deployment and infrastructure myself.
EDUCATION
Concordia University B.Sc. Computer Science
COURSES
Artificial Intelligence Machine Learning Web Programming OS, Data Structures Algorithms & Analysis
LOCATION
Montreal, Canada
INTERESTS
Systems Design Competitive Gaming Cooking and Sketching
▶ FEATURED PROJECTS ◀
Flâneur
Flâneur is a full stack travel discovery platform that helps users find hidden gem activities globally using interactive maps and advanced filtering. Built the backend with FastAPI deployed on AWS EC2, handling complex geospatial queries for real-time location-based data. Designed a responsive UI with React and Tailwind CSS, and implemented secure API endpoints with proper validation against SQL injection.
MovieFinds
MovieFinds is a full stack movie discovery platform that allows users to search, explore, and manage movies. Built the backend API, handled user-related functionality, and connected the system to a frontend interface. This project focused on backend development, API communication, and building applications where backend logic drives the user experience.
Spatial GATv2 - Airbnb Demand Prediction
Graph neural network to classify high-demand Airbnb listings in Montreal. Built a three-layer Spatial GATv2 network (GATv2Conv, Brody et al. 2022) representing listings as nodes with spatial edges weighted by geographic proximity. Benchmarked against Logistic Regression and Gradient Boosting across accuracy, F1, and ROC-AUC, framing Gradient Boosting's outperformance as a publishable finding on GNN limits in feature-dominant regimes. Produced interactive Folium map visualizations of the listing graph.
Deep Learning for Dermatological Diagnosis
Group project (5 members) applying ResNet-50, VGG-16, and MobileNetV2 to skin lesion classification across HAM10000 and ISIC 2019 datasets. Owned the data pipeline: cleaning, 4-class selection, and stratified 70/15/15 train/val/test splitting across ~7,500 images. Built shared evaluation scripts (accuracy, F1, ROC-AUC, confusion matrices), Grad-CAM interpretability figures, and a hyperparameter grid-search runner over learning rate and batch size.
▶ INVENTORY ◀
Languages
Python, JavaScript, TypeScript, Java, SQL, HTML/CSS
AI / ML
PyTorch, PyTorch Geometric, scikit-learn, NumPy, pandas, GATv2, CNNs, NLP, Grad-CAM
Frontend
React, Next.js, React Native (Expo), Tailwind CSS
Backend
FastAPI, Node.js
Infrastructure
AWS, Docker, Kubernetes, PostgreSQL, Vercel, Render, Git
Tools
Claude Code, Cursor, Linux, Figma, Procreate
▶ GET IN TOUCH ◀
Have a project in mind? Want to collaborate? Just want to say hi? Feel free to reach out!
I'm always open to new opportunities and interesting conversations.
Whether it's a quick question or a full project proposal, I'm here to help!