Khaled is currently unavailable
Khaled Abdelhamid
Verified Expert in Engineering
Machine Learning Engineer and Developer
6th of October City, Giza Governorate, Egypt
Toptal member since June 20, 2022
Bio
Khaled is a senior machine learning engineer with four years of experience building state-of-the-art solutions. He is passionate about natural language processing and computer vision. Khaled specializes in web scraping, data collection, and publishing competitive datasets in the Arabic language.
Portfolio
Agolo
AI Design, Machine Learning, Natural Language Processing (NLP)...
Safeguard My Car
Artificial Intelligence (AI), Chatbots, Chatbot Conversation Design...
Udacity
Python 3, Data Pipelines, Advisory
Experience
- Python 3 - 4 years
- PyTorch - 3 years
- Natural Language Toolkit (NLTK) - 3 years
- SpaCy - 3 years
- TensorFlow - 3 years
- Amazon Web Services (AWS) - 3 years
- Docker - 2 years
- Rasa NLU - 2 years
Work Experience
Machine Learning/NLP Engineer
2022 - PRESENT
Agolo
- Led the implementation of multiple cutting-edge machine learning models for natural language processing (NLP) applications, achieving state-of-the-art accuracy and performance.
- Developed custom pre- and post-processing pipelines to enhance named-entity recognition (NER) robustness and accuracy while also improving multilingual support.
- Contributed to the development and optimization of a knowledge base system specifically tailored for information retrieval, streamlining access to critical information within the organization.
- Created tailored evaluation pipelines to meticulously measure the accuracy of various machine learning models, contributing to ongoing excellence.
- Constructed comprehensive quantitative evaluation dashboards, empowering development teams to conduct rigorous assessments and identify areas for improvement.
- Actively engaged in diverse research initiatives aimed at harnessing the capabilities of large language models, integrating them into the pipeline using tools like LangChain.
- Orchestrated end-to-end pipelines to facilitate seamless communication among numerous interconnected services, optimizing system performance.
- Played a pivotal role in seamlessly integrating machine learning models into CI/CD pipelines, ensuring scalability and enhanced performance.
- Devised and executed extensive performance stress testing protocols for deployed ML services, determining optimal resource requirements under varying loads.
- Managed a dynamic team of five professionals, successfully achieving the goal of implementing robust multilingual support within an impressive timeframe of under a month.
Technologies: AI Design, Machine Learning, Natural Language Processing (NLP), OpenAI GPT-3 API, Large Language Models (LLMs), Data Cleansing, OpenAI GPT-4 API, Automatic Speech Recognition (ASR), Text-to-Speech (TTS), LangChain, Serverless, SQL, Retrieval-augmented Generation (RAG), Vector Databases, Generative Artificial Intelligence (GenAI), Hugging Face, Machine Learning Operations (MLOps), OpenAI
AI Expert
2023 - 2024
Safeguard My Car
- Developed a large language model (LLM)-based sales voice chatbot integrated with the client database, enabling real-time, high-throughput client interactions with an average response latency of one second.
- Implemented a retrieval-augmented generation (RAG) system to address user FAQs during live calls, utilizing LangChain and ChromaDB for seamless support.
- Directed project development using Jira and Agile methodologies, overseeing sprint coordination for efficient task management. Led a team of five engineers to optimize the architecture, ensuring market readiness.
Technologies: Artificial Intelligence (AI), Chatbots, Chatbot Conversation Design, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), OpenAI GPT-3 API, OpenAI GPT-4 API, Speech-to-Text (STT), Automatic Speech Recognition (ASR), Text-to-Speech (TTS), LangChain, Serverless, Model Tuning, Retrieval-augmented Generation (RAG), Vector Databases, Generative Artificial Intelligence (GenAI), Hugging Face, Machine Learning Operations (MLOps), OpenAI
Session Lead in Computer Science
2022 - 2023
Udacity
- Successfully managed weekly sessions with a cohort of 35 students, achieving an impressive graduation rate of 93%.
- Demonstrated exceptional leadership as a top-rated session lead, consistently earning a 5-star rating and receiving outstanding feedback from students.
- Actively maintained and provided mentorship to numerous students through the Slack platform, fostering a supportive and conducive learning environment.
Technologies: Python 3, Data Pipelines, Advisory
Machine Learning Engineer
2019 - 2022
Online Freelance Agency
- Contributed to building machine learning-based solutions for numerous customers.
- Designed and established data pipelines for NLP-based projects.
- Acted as the machine learning freelance achieving the maximum ratings and a 98% job success rate.
Technologies: Python 3, Natural Language Toolkit (NLTK), SpaCy, NumPy, Pandas, Git, Rasa NLU, PyTorch, TensorFlow, Keras, Amazon Web Services (AWS), Docker, MATLAB, Deep Learning, Machine Learning, Data Science, PostgreSQL, TensorBoard, BERT, DeepSpeech, Statistics, Visualization, A/B Testing, Matplotlib, Seaborn, Streamlit, Amazon S3 (AWS S3), Amazon EC2, Amazon Elastic Block Store (EBS), SciPy, Jupyter, Jupyter Notebook, OpenCV, LabVIEW, Scrapy, Beautiful Soup, Selenium, Linux, Windows, Visual Studio Code (VS Code), Slack, Trello, Notion, English, Scikit-learn, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Regex, Optical Character Recognition (OCR), Data Visualization, Google Cloud Platform (GCP), Python Asyncio, LaTeX, Web Dashboards, REST, Computer Vision, Bash, Bash Script, FFmpeg, Web Scraping, Google Colaboratory (Colab), Rasa.ai, Named-entity Recognition (NER), Chatbots, Image Processing, Markdown, YAML, JSON, ELK (Elastic Stack), Data Cleaning, Data Collection, FastAPI, Data Analysis, Data Processing, Signal Processing, Digital Signal Processing, Artificial Intelligence (AI), Text Analytics, Python, GPU Computing, Cloud, Machine Vision, Neural Networks, Apache Spark, ETL, Text Generation, Language Models, Speech Recognition, APIs, Data Pipelines, Google BigQuery, AI Design, Chatbot Conversation Design, Project Consultancy, Advisory, ChatGPT, OpenAI GPT-3 API, Large Language Models (LLMs), Azure Machine Learning, Amazon Machine Learning, Google Cloud Machine Learning, Data Cleansing, OpenAI GPT-4 API, LlamaIndex, LangChain
Research Assistant
2020 - 2021
Zewail City
- Developed a transformer-based model for the Arabic text diacritization task and outperformed the state-of-the-art method with a total accuracy of 98%.
- Processed the Arabic speech data and coded a deep learning-based model to do speech-to-text over the processed data.
- Wrote comparative articles and literature reviews on natural language processing and deep learning and their application in Arabic for non-specialized Arabic readers.
Technologies: Python 3, TensorBoard, TensorFlow, Deep Learning, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), SpaCy, Natural Language Toolkit (NLTK), NumPy, Pandas, Git, Rasa NLU, PyTorch, Keras, Amazon Web Services (AWS), Docker, MATLAB, Machine Learning, Data Science, BERT, DeepSpeech, Statistics, Visualization, A/B Testing, Matplotlib, Seaborn, Streamlit, Amazon S3 (AWS S3), Amazon EC2, Amazon Elastic Block Store (EBS), SciPy, Jupyter, Jupyter Notebook, OpenCV, LabVIEW, Linux, Windows, Visual Studio Code (VS Code), Slack, Trello, Notion, English, Scikit-learn, Regex, Optical Character Recognition (OCR), Data Visualization, Google Cloud Platform (GCP), LaTeX, Web Dashboards, REST, Computer Vision, Bash, Bash Script, FFmpeg, Web Scraping, NodeMCU, Google Colaboratory (Colab), Named-entity Recognition (NER), Chatbots, Image Processing, Markdown, YAML, JSON, Data Cleaning, Data Collection, FastAPI, Data Analysis, Data Processing, Signal Processing, Digital Signal Processing, Artificial Intelligence (AI), Text Analytics, Python, GPU Computing, Neural Networks, ETL, PySpark, Speech Recognition, APIs, Data Pipelines
Machine Learning Engineer
2020 - 2020
Proteinea (startup)
- Conducted training, testing, and logging pipelines to use deep learning in predicting protein expression levels using DNA features.
- Established monitoring systems of the factory production status for logging and further analysis.
- Developed a computer vision-based system to count moving occluded objects over a conveyor built with fast and robust predictions.
Technologies: Python 3, PyTorch, TensorFlow, Pandas, NumPy, Scikit-learn, SciPy, Amazon Web Services (AWS), TensorBoard, Data Science, Git, Keras, Docker, Deep Learning, Machine Learning, Statistics, Visualization, A/B Testing, Matplotlib, Seaborn, Amazon S3 (AWS S3), Amazon EC2, Amazon Elastic Block Store (EBS), Jupyter, Jupyter Notebook, OpenCV, Linux, Windows, Visual Studio Code (VS Code), Slack, Trello, Notion, English, Optical Character Recognition (OCR), Data Visualization, LaTeX, Web Dashboards, REST, Computer Vision, Bash, Bash Script, NodeMCU, Google Colaboratory (Colab), Image Processing, Markdown, JSON, Data Cleaning, Data Collection, Data Analysis, Data Processing, Signal Processing, Digital Signal Processing, Artificial Intelligence (AI), Text Analytics, Python, GPU Computing, Cloud, Machine Vision, Neural Networks, ETL, APIs, Data Pipelines, AI Design
Project History
Event Information Extraction from Tweets for Tech Startup
The project aims to extract information like dates, phone numbers, organizations, etc., from user tweets and save them in a structured format.
I have implemented the project using a combination of pattern recognition techniques using regex in addition to using Spacy's named entity recognition features. As for the dates, I have used a temporal model to parse the text and extract the dates and normalize them to be queried and analyzed.
I have implemented the project using a combination of pattern recognition techniques using regex in addition to using Spacy's named entity recognition features. As for the dates, I have used a temporal model to parse the text and extract the dates and normalize them to be queried and analyzed.
Ranking and Associating Job Descriptions Along With Their Job Families for HR Agency
The project aims to get the similarity between each job description along with a job family description and level. Using NLP, I was able to make a ranking algorithm that gives a similarity score between each job and its job family.
Text Mining and Analysis for Marketing Company
The project aims to build multiple APIs over audio dialogues of focus groups and market surveys. The required modules included entity recognition, speech-to-text, text summarization, translation, key phrase extraction, and sentiment analysis.
End-to-end Chatbot for Customer Services
https://www.youtube.com/watch?v=SJRmzDfWIecI developed a voice chatbot using the Rasa framework that contains the following features:
• Connects to REST APIs to handle users' voice and text commands from their database.
• Performs the extraction of the answers from archived business contracts using deep learning and fuzzy search.
• Conducts named-entity recognition, intent classification, and form-filling to extract all the needed information to perform a successful information retrieval for users.
• Connects to REST APIs to handle users' voice and text commands from their database.
• Performs the extraction of the answers from archived business contracts using deep learning and fuzzy search.
• Conducts named-entity recognition, intent classification, and form-filling to extract all the needed information to perform a successful information retrieval for users.
AI-based OCR for Extracting Data from Electric Meters
The project aimed to build an AI OCR model and extract text and numbers from a calculator-type digital screen. The provided dataset was used to train and extract details with nearly 100% accuracy, while the setup was in a controlled environment indoors with an exact camera shot.
First, I applied gamma correction to the images and other noise-canceling techniques to reach the optimal state the OCR engine could handle. Then, I developed the solution using EasyOCR and PaddleOCR. To enhance the performance of the seven-segment display, I created a dataset using the TRDG library with custom seven-segment fonts to train the OCR engines. Because there were expected text patterns, I used NLP methods to correct text based on the most likely predictions.
First, I applied gamma correction to the images and other noise-canceling techniques to reach the optimal state the OCR engine could handle. Then, I developed the solution using EasyOCR and PaddleOCR. To enhance the performance of the seven-segment display, I created a dataset using the TRDG library with custom seven-segment fonts to train the OCR engines. Because there were expected text patterns, I used NLP methods to correct text based on the most likely predictions.
YouTube Scraper
I architected highly parallel software to scrape channel video information and its timestamp. I used Selenium to scrape dynamic web pages, retrieve comments containing video timestamps detected with the regular expression, and deployed the system on an EC2 instance with continuous monitoring.
Traffic Analysis with CCTV Cameras
The goal was to classify and track all the moving objects in a fixed traffic spot to analyze the traffic movement. The video feed came from fixed-angle CCTV cameras with almost 12 hours of coverage for each location. The data was chunked and processed into reasonable sizes to be fed inside the model sequentially.
I used bash scripting and FFmpeg for the video processing and the FairMOT model to classify and track the objects. The categories were pedestrians, cars, motorcycles, tricycles, and buses. The tracking data was filtered and used to create a traffic flow map, a visualization chart showing the most condensed areas in a specific location for a given time.
I used bash scripting and FFmpeg for the video processing and the FairMOT model to classify and track the objects. The categories were pedestrians, cars, motorcycles, tricycles, and buses. The tracking data was filtered and used to create a traffic flow map, a visualization chart showing the most condensed areas in a specific location for a given time.
Email Scraper
The project aimed to extract emails from 30,000 URLs saved in a CSV file.
I used Scrapy to make asynchronous requests to speed up scraping. For each page, I extracted all the emails using regular expression patterns to ensure the minimum amount of false negatives. The code finished with successful results and extracted emails from 90% of the websites.
I used Scrapy to make asynchronous requests to speed up scraping. For each page, I extracted all the emails using regular expression patterns to ensure the minimum amount of false negatives. The code finished with successful results and extracted emails from 90% of the websites.
Genuine Artistic Images from Custom Tags Using GANs
The project aimed to generate artistic images from royalty-free images using deep learning. The model inputs were name tags or categories such as cats, cars, and animals, and the expected output was an artistic representation of the given image with unique features.
I built a Streamlit dashboard to interact smoothly with the user and get the name tags. Then I scraped websites containing high-resolution free images and used the DeepDream model in processing the scraped images to get newer artistic ones that were unique and related to the given name tags. The output was delivered, allowing the user to refresh and edit the parameters provided to the DeepDream model.
I built a Streamlit dashboard to interact smoothly with the user and get the name tags. Then I scraped websites containing high-resolution free images and used the DeepDream model in processing the scraped images to get newer artistic ones that were unique and related to the given name tags. The output was delivered, allowing the user to refresh and edit the parameters provided to the DeepDream model.
Named-Entity Extractor Pipeline
The project included processing a dataset of sentences and labeling all the ones that contained names of locations. Once I extracted the entities, the project required further processing to standardize the writing of the extracted places and get detailed information about the location, such as the country and city.
Text Data Augmentation
The goal was to perform text data augmentation on a given dataset. I used a combination of the following techniques:
• random replacements of characters in words based on the most likely mistakes for such a word;
• use of synonyms and antonyms with negation;
• translation of phrases into a different language and then re-translation to the original language (this method could chain multiple languages sequentially); and
• paraphrasing sentences using the BERT-based model.
• random replacements of characters in words based on the most likely mistakes for such a word;
• use of synonyms and antonyms with negation;
• translation of phrases into a different language and then re-translation to the original language (this method could chain multiple languages sequentially); and
• paraphrasing sentences using the BERT-based model.
Converting Driver License Images Into Tabulated Data
The project aimed to extract all the possible information from a driver's license image and make a structured dataset with all the associated records. I used an OCR engine to scrap all the data from the cards and associate each text block into a category, such as a name, ID, or expiration date. The records were saved into a CSV, and I coded a Streamlit dashboard to show the dataset and analyzed the dataset further, including age distribution and card locations.
ECG Analysis and Cardiac Arrhythmia Detection
The project classified cardiac arrhythmia cases using the reading of ECG signals.
I built hardware to extract the ECG signal from a person using NodeMCU and ECG custom kit. Then I read and uploaded the data into Firebase and trained and deployed a machine learning classifier in GCP. The signal could have six categories—normal and the other five cardiac arrhythmia types. The signal was processed and analyzed with LabVIEW.
I built hardware to extract the ECG signal from a person using NodeMCU and ECG custom kit. Then I read and uploaded the data into Firebase and trained and deployed a machine learning classifier in GCP. The signal could have six categories—normal and the other five cardiac arrhythmia types. The signal was processed and analyzed with LabVIEW.
Website Topic Categorization
The goal was to label a given list of websites in a category based on the content, such as sports, economy, and news. The URLs were delivered in a CSV file.
I used Scrapy to process all the links asynchronously and get all the paragraph and header tags, which contained most of the informative text inside the URL. Then I used Google NLP API to extract the most likely category, saved the data continuously, and stored it in another CSV file.
I used Scrapy to process all the links asynchronously and get all the paragraph and header tags, which contained most of the informative text inside the URL. Then I used Google NLP API to extract the most likely category, saved the data continuously, and stored it in another CSV file.
JPEG Image Compression
https://github.com/Khaled-Abdelhamid/JpegThe project aimed to implement the basic JPEG image compression algorithm and develop metrics to monitor the data loss and compression efficiency.
I was able to get off 50% compression efficiency with almost 0.02% data loss between the raw and compressed image.
I was able to get off 50% compression efficiency with almost 0.02% data loss between the raw and compressed image.
Spectrum Analyzer Application
https://github.com/Khaled-Abdelhamid/Spectrum-analyzer-applicationThe project involved building an application that performed spectrum analysis functionalities over a given audio signal like FFT, filtering, and windowing. The application also provided the user with an option to test and compare a variety of arbitrary signals.
Education
2016 - 2021
Bachelor of Engineering Degree in Communications and Information Technology Engineering
University of Science and Technology, Zewail City - Giza, Egypt
Certifications
APRIL 2021 - PRESENT
Natural Language Processing Specialization
Coursera
MARCH 2021 - PRESENT
AWS Cloud Practitioner Essentials
Amazon Web Services (AWS)
MARCH 2021 - PRESENT
Advanced Data Analysis Nanodegree Program
Udacity
NOVEMBER 2020 - NOVEMBER 2022
IELTS | 7.5 | C1
British Council
SEPTEMBER 2019 - PRESENT
Deep Learning Specialization
Coursera
Skills
Libraries/APIs
Natural Language Toolkit (NLTK), SpaCy, NumPy, Pandas, Rasa NLU, PyTorch, SciPy, TensorFlow, Keras, Matplotlib, OpenCV, Beautiful Soup, Scikit-learn, PySpark, DeepSpeech, Python Asyncio, FFmpeg
Tools
Git, Jupyter, Notion, Named-entity Recognition (NER), Seaborn, LabVIEW, Slack, Trello, Rasa.ai, ChatGPT, Kibana, Azure Machine Learning, MATLAB, TensorBoard, Amazon Elastic Block Store (EBS), LaTeX, Plotly, ELK (Elastic Stack), Jira
Languages
Python 3, Regex, Markdown, Python, C++, C, SQL, Julia, C#, Bash, Bash Script, YAML
Frameworks
Streamlit, Hadoop, Spark, Scrapy, Selenium, Apache Spark, LlamaIndex
Platforms
Jupyter Notebook, Amazon Web Services (AWS), Docker, Visual Studio Code (VS Code), Amazon EC2, Linux, Windows, Google Cloud Platform (GCP), Firebase
Storage
JSON, Data Pipelines, PostgreSQL, Amazon S3 (AWS S3), Elasticsearch, HBase, SQL Server 2016, Apache Hive
Paradigms
REST, ETL
Other
Deep Learning, Machine Learning, Data Science, English, Natural Language Processing (NLP), Google Colaboratory (Colab), Chatbots, Image Processing, Data Cleaning, Data Processing, Data Analysis, Artificial Intelligence (AI), Text Analytics, Data Engineering, Neural Networks, Speech Recognition, APIs, AI Design, Chatbot Conversation Design, Project Consultancy, Advisory, Generative Pre-trained Transformers (GPT), Large Language Models (LLMs), Data Cleansing, LangChain, Retrieval-augmented Generation (RAG), Vector Databases, Generative Artificial Intelligence (GenAI), Hugging Face, OpenAI, BERT, Data Visualization, Web Dashboards, Computer Vision, Web Scraping, FastAPI, Signal Processing, Digital Signal Processing, GPU Computing, Cloud, Machine Vision, Text Generation, Language Models, Google BigQuery, OpenAI GPT-3 API, Amazon Machine Learning, Google Cloud Machine Learning, OpenAI GPT-4 API, Speech-to-Text (STT), Automatic Speech Recognition (ASR), Text-to-Speech (TTS), Serverless, Model Tuning, Machine Learning Operations (MLOps), Statistics, Visualization, A/B Testing, Optical Character Recognition (OCR), NodeMCU, Data Collection, Speech to Intent
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring