Abdullah Usama
Software Engineering Student & ML/AI Engineer
Passionate about machine learning, computer vision, and building scalable software solutions. Currently exploring AI Agents and Automation.
About
I'm a final-year Software Engineering student at the National University of Sciences and Technology (NUST), with a experience in AI Agents, Automations and Web Development.
I have worked on fine-tuning Large Language Models (LLMs) and Vision-Language Models (VLMs), developing intelligent AI agents, creating automation solutions, and building scalable web applications. I'm passionate about using cutting-edge AI technologies to solve real life problems.
Education
Bachelor of Software Engineering
National University of Sciences and Technology (NUST)
School of Electrical Engineering and Computer Science (SEECS)
Focus Areas
Machine Learning & Computer Vision
Full-Stack Web Development
AI-Powered Applications
Experience
June 2025 – Present
AI Intern
Crimson Labs, SEECS
- - Working on finetuning LLMs
- - Creating AI Chatbots for Education
- - RAG - Based Systems for Learning
April 2025 – May 2025
ML Intern
OneScreen Solutions, San Diego, California (Remote)
- - Worked with Vision Transformers (ViT-32) and Vision-Language Models (VLMs) like PaLI-Gemma
- - Achieved 7-10% higher mAP by reducing label noise and improving object localization
- - Developed end-to-end pipeline combining SAM's pixel-level masks with YOLO annotations
June 2024 – Aug 2024
Computer Vision Intern
Machine Vision & Intelligent Systems Lab (MachVis), SEECS
- - Engineered computer vision pipelines for real-time object detection and tracking
- - Implemented robust feature extraction methods including ORB and optical flow
- - Developed real-time tracking systems using SORT and Kalman Filters
Selected Projects
The Pakistan News AI Assistant is an intelligent agent designed to help students and aspirants understand editorial and opinion articles from DAWN (Editorial & Op-Ed), The Tribune (Editorial), and ParadigmShift (National & International Relations) newspapers. It provides various functionalities, including getting all the articles related to a certain topic, scraping articles, extracting key information.
Pipeline to make YOLO bounding boxes more precise and tight around the target objects. Making the data more precise and improving localization.
A fine-tuned Mistral-7B-Instruct-v0.3 model capable of generating opinion-style text in the distinctive writing style of Pakistani diplomat, journalist, and political scientist, Maleeha Lodhi.
Comprehensive football match analysis system with player tracking, distance estimation, and possession analysis.
Full-stack e-commerce platform with secure authentication, multilingual support, and payment integration.
Real-time hands-free volume control system using dynamic hand gestures, leveraging MediaPipe and Pycaw for precise landmark detection and system audio integration.
Cloud-based video streaming application with microservices architecture, secure authentication, scalable backend, and real-time media processing on Google Cloud.
Educational implementation of the Transformer architecture based on the 'Attention Is All You Need' paper. Built from scratch in PyTorch to understand attention, positional encoding, multi-head mechanisms, and encoder-decoder structure.
Skills & Technologies
Programming Languages
JavaScript
Python
C++
SQL
Web Development
React.js
Next.js
Tailwind CSS
TypeScript
Node.js
Express.js
FastAPI
RESTful APIs
AI/ML & Computer Vision
TensorFlow
OpenCV
YOLO
MediaPipe
LangChain
LangSmith
Databases
MongoDB
MySQL
PostgreSQL
Cloud & Deployment
Docker
Vercel
Render
Google Cloud
Tools & Platforms
Git
GitHub
Clerk.com
Stripe
Pycaw
React-i18n
Let's Connect
I'm always interested in discussing new opportunities, collaborations, or innovative projects. Feel free to reach out if you'd like to connect.