Umair Amir

I am a Computer Science graduate from LUMS with a dual specialization in Full Stack Engineering and Deep Learning. My technical background allows me to handle both scalable web architectures and complex computer vision research pipelines.

In my recent role as a Software Engineer at Parcel To Ship, I engineered a full-stack shipping marketplace using React.js, Node.js, and SQL. I went beyond basic CRUD applications by deploying the platform under a microservices-based Control Tower model to improve system reliability and reduce deployment downtime. My work involved robust backend development, including integrating Stripe for secure transactions and connecting courier APIs like Aramex and iMile for real-time logistics tracking. Additionally, I have mobile development experience from building a fleet management system in React Native, where I integrated Google Location API for trip logging and implemented role-based dashboards for administrative oversight.

On the AI engineering front, I serve as a Research Assistant at the CV & Graphics Lab and Center for Speech and Language Technologies Lab (CSALT), where I design advanced architectures integrating Mixture of Experts (MoE) with Swin Transformers. I engineer end-to-end pipelines in PyTorch to standardize data preprocessing for high-dimensional visual data. My portfolio includes developing a Forest Fire Detection framework that combines Swin Transformers with Faster R-CNN detectors to improve accuracy across satellite and outdoor datasets. I also applied these skills to medical imaging, building a deep learning pipeline with Enhanced Squeeze-and-Excitation modules to classify Alzheimer's MRI slices with high precision.

My expertise extends to Natural Language Processing and hardware-aware optimization. I recently investigated system bottlenecks in on-device LLM fine tuning, analyzing flash write endurance on edge hardware to propose sustainable deployment strategies. I also built an Urdu News Classification system using web scraping and custom text preprocessing pipelines to handle low-resource language constraints.

I am proficient in Python, C++, TypeScript, and JavaScript. My core stack includes React.js, Next.js, FastAPI, MongoDB, Supabase, and SQL. I am available for complex development projects that require rigorous engineering standards.