M_ELSHERIF
AI / ML Engineer — Assiut, Egypt

MARIAM
ELSHERIF

fi hybrid trackers work, elihw compilers read backwards — I build ML systems that ship, not just notebooks.

CS student and NASA Space Apps Global Finalist with hands-on experience across computer vision, deep learning, and full-stack AI — from aerial object trackers to a self-designed programming language with mirrored syntax.

View Work → Get In Touch
3.62CGPA · CS, Assiut National University
3Shipped end-to-end ML systems
NASASpace Apps Global Finalist
✦ 01 / SELECTED WORK
Projects
Three systems, three problems: tracking under motion, a language that mirrors itself, and signals from other stars.
PROJECT 01APR 2026

HybridTrack

Aerial Single-Object Tracker
  • Hybrid architecture fusing MobileNetV3-Large coarse proposals with EfficientViT-M2 semantic features via a custom lightweight cross-attention module.
  • Progressive backbone unfreezing, cosine window + scale penalty at inference, and exponential box smoothing to stay locked on under fast motion and occlusion.
PyTorchComputer VisionEfficientViTAttention
PROJECT 02FEB – APR 2026

Merror

Mirrored Programming Language
  • Compiled language with mirrored syntax (fi/elihw instead of if/while) — variables, functions, conditionals, loops, built from scratch in Python.
  • Full pipeline: regex lexer, recursive-descent parser, semantic analysis, and codegen transpiling .mr files to C++.
  • Shipped an interactive web playground to write and run Merror in the browser, no local setup.
PythonCompilersC++Web
PROJECT 03OCT – NOV 2025
NASA Space Apps Global Finalist

AEXIS

Transit-Method Exoplanet Prediction
  • Full-stack platform (React + Node.js) classifying planetary signals from real NASA Kepler/TESS data, with interactive exploration.
  • Dual models: CNN for light-curve image classification, LSTM for time-series transit detection.
  • Role-based UX — researchers fine-tune and export, enthusiasts get a one-click demo.
React.jsNode.jsCNNLSTM
✦ 02 / TRACK RECORD
Experience
Internships spanning speech ML, LLM evaluation, and data storytelling.
Oct 2025 — Dec 2025

AI/ML Engineer Intern

Voxa · Remote
  • Built a speech-to-text pipeline transcribing user recordings and extracting structured features for downstream analysis.
  • Developed a feedback generation module producing personalized, actionable coaching on delivery and performance.
  • Evaluated 6 LLM candidates on emotional-intelligence benchmarks and integrated the selected model into production.
Apr 2025 — Jun 2025

Data Science Intern

ESSR · Assiut, Egypt
  • Cleaned and analyzed 1,200+ participant records, improving outreach efficiency by 25%.
  • Built interactive dashboards tracking workshop performance across 200+ attendees.
  • Produced 20+ visual reports used to evaluate and communicate program outcomes.
Currently: Interning at Ejada's Customer Experience Solutions Practice, building Dynamics 365 / Power Platform solutions.
✦ 03 / TOOLKIT
Skills
Languages, ML/DL frameworks, web, and infrastructure.
Languages
CC++PythonJavaSQLC#
ML & Deep Learning
PyTorchTensorFlowScikit-learnOpenCVHuggingFace Transformers
Web
React.jsNode.jsFlaskASP.NET MVC Core
Data & Visualization
PandasNumPyMatplotlibStreamlit
Infrastructure
DockerGit & GitHubMySQLPostmanSeleniumpytest
Design
FigmaFramer