Ahmed Muhammad

Experience

• 

  ZTechnium

  Jan 2020 - Mar 2020

  Machine Learning Intern

  - Learned and deployed machine learning models and algorithms.- Implemented text classification model, recognition system and Sentiment analysis model in English sentence.

• 

  Wakeb_Data

  Jan 2020 - Mar 2020

  Machine Learning Intern

  - Learned and deployed machine learning models and algorithms and Worked on chat-bots and the dialects in Arabic language.- Implemented sentiment analysis model on Arabic sentences, thet analysis of the content of social networks into negative and positive impressions.

• 

  M06

  Apr 2020 - Jun 2020

  Remote Machine Learning Engineer

  - Worked on automated machine learning (AutoML) models.- Implemented text classification model and sentiment analysis model.- Get ~97% accuracy on classified English movies reviews by using BERT model.- Worked on resume parsing by used 2 models (BERT and custom Spacy NLP).

• 

  Cloudypedia

  Aug 2021 - Nov 2021

  Remote Data Scientist Cloud

  - Work on Cloud computing specially on Google Cloud Platform (GCP).- Worked on Analysis a huge data on BigQuery.- Implemented Machine learning Algorithms on huge data.

• 

  Wakeb_Data

  Nov 2021 - now

  Data Scientist

  Deep Fake For Arabic Text. Working on Generative Adversarial Networks (GANs) for text generation andclassification text. The project aims to generate human-like text using GANs and build a classification model capable of distinguishing between text generated by machines and human-generated text.Building a Transformer-based model from scratch for text generation. The generator model's encoder will be repurposed to construct a classification model by adding a Linear layer with a Sigmoid function will augment the encoder to perform binary classification.- Working On Object detection and tracking algorithms, designed and implemented Multi Object Tracking based on colors features from histogram of image and deep features from Resiudal Neural Network then applied weighing sum to mixed the colorful and deep features values then pass them to assign methods (distance, IOU and drop boxes every step) to get the best matched track-ID between last tracks. Improved the accuracy of the system by throwing every K frame to the detection model, not every frame.- Implemented batch detection on real time streaming to trace and track people and calculated the speed of cars, working on 3 Cameras in the same time and streaming it using the StreamLit system takes 25 frames per seconds with accuracy 96%.- Working on Graph Neural Network algorithms (Connected Component, Convolutional Neural Network) and using networkX to analyze, manipulate and visualize Graph Neural Network.- Working on Embedding images from satellites and drones by using Vision Transformer from Hugging Face and merged it with Resiudal Neural Network to get complex patterns in images to calculate similarities of images.- Working on Image Processing, Machine Learning and Deep Learning Algorithms, Using SIFT to get key points descriptors in images and applying similarity . Show less

• 

  CYNERV

  Jan 2023 - now

  AI Team Lead