Signal processing and GNSS
Algorithms that turn raw, noisy measurements into reliable positions and insight. My current day-to-day, in Python, MATLAB and C/C++.
I'm Mariam, an Emirati software engineer in Abu Dhabi. I build algorithms in signal processing, AI and GNSS, and I like turning hard problems into clean, working code.
Open to freelance projects
I studied computer engineering at Khalifa University, then went deeper into artificial intelligence for my master's, where I researched brain-inspired spiking GANs that help underwater robots see through murky water. Along the way I found the problems I enjoy most sit where math meets real-world signals: noisy data, tight constraints, and answers that have to be right.
Today I work as a software engineer in the GNSS and defense sector, and I'm steadily growing into space technology. Satellites, orbits and the signals they send down are what pull me forward, and in 2024 I trained in remote sensing and geospatial analytics with the UAE Space Agency. Before that I helped turn Jais, the Arabic large language model, into a product at MBZUAI. Outside work I keep building side projects, learning new tools, and improving things that could work better.
Underwater, light scatters and fades fast, so cameras see a blue-green haze. That makes navigation hard for underwater robots. I trained GAN-based dehazing models on underwater datasets and built them with spiking neural networks, the brain-inspired models that run efficiently on neuromorphic hardware. The goal is clearer images in real time, using little power.
Illustration only, drawn in the browser. Not output from the thesis model.
Algorithms that turn raw, noisy measurements into reliable positions and insight. My current day-to-day, in Python, MATLAB and C/C++.
Spiking and generative models for computer vision, Arabic NLP from diacritization to text generation, and forecasting for smart homes. Models that do something useful.
Genetic algorithms, swarm optimization, hashing schemes and graph ranking. I enjoy making the hard search fast.
Each one is on GitHub. Filter by the kind of problem.
Aligning biological sequences with a genetic algorithm and Tuna Swarm Optimization, benchmarked on BAliBase and scored with sum-of-pairs, entropy and gap-penalty metrics.
Generating Arabic text with AraGPT2 transformer models, with Farasa and PyArabic preprocessing.
My senior design project for the KU smart home at Solar Decathlon Middle East, which placed 2nd in energy efficiency. It forecasts solar output and power use, and recommends when to run appliances, through a Xamarin app.
A hybrid genetic and ant colony approach that schedules classes, rooms and instructors without clashes, backed by SQLite.
A Flask web app for Python learners, with an in-browser IDE, project notes and a reporting dashboard.
A mental-math reaction game with an LCD, keypad, IR sensor, RGB LED and speaker. Featured on Arduino Project Hub.
Both collision strategies implemented and compared head to head, plus a perfect-hashing companion project.
Built for the Digicode-Ed Challenge at Coders HQ: a ChatGPT-powered tool that helps teachers create lesson plans tailored to their students.
Building and maintaining software tools that automate technical workflows for GNSS projects.
Trained in geospatial analytics and remote sensing through a UAE Space Agency initiative, working on practical space projects.
Helped turn Jais, the leading Arabic large language model, into a real product.
Ran labs for COSC-101 with 50+ students and assisted on COSC-434, Intro to Machine Learning.
Built RestEasy, a cross-platform app prototype in C# and Xamarin.Forms that monitors children's nighttime asthma.
Hands-on training in geospatial analytics and remote sensing: turning satellite imagery and Earth-observation data into answers, through practical space projects.
Advanced training on Skydel, Safran's software-defined GNSS simulator for generating and testing satellite navigation signals.
AI course with the UAE Office of AI and the National Program for Coders.
Immersive Arabic NLP program, building toward an Arabic diacritizer.
Four intensive weeks of C, shell, git and algorithms, with daily projects and exams.
Spreadsheets, SQL, R and data visualization.
Six-week hands-on training.
Specialization on building and using AI responsibly.
Always playing somewhere nearby.
Asking why, then asking it again.
Stories, framing, and a good soundtrack.
Picking up something new on purpose.
If it can work better, I want to try.