Ibrahim Mahmoud Ahmed, Ph.D.
Verified Expert in Engineering
Machine Learning Developer
Perth, Western Australia, Australia
Toptal member since April 22, 2019
Ibrahim is a veteran data scientist and software developer with a passion for deep learning and extensive experience in statistics and time series. Ibrahim is interested in helping businesses take full advantage of their data and take data-driven actions. His principle is to observe, model, and corroborate, disrupting the status quo with artificial intelligence.
Portfolio
Experience
- SQL - 20 years
- Data Science - 10 years
- Python 3 - 10 years
- Statistical Learning - 10 years
- Deep Learning - 5 years
- TensorFlow - 5 years
- ArcGIS - 5 years
- PyTorch - 2 years
Availability
Preferred Environment
Jupyter Notebook, Google Cloud Platform (GCP), Azure, Amazon Web Services (AWS), ChatGPT, Bloomberg, OpenAI GPT-3 API, OpenAI GPT-4 API, Selenium, 3D Image Processing
The most amazing...
...berthing project I've developed for a mining company used SplatNet, an enabled hierarchical and spatially aware feature detection to be used by AR.
Work Experience
AI developer for a project in the healthcare industry
Quilez and Associates Inc
- developed chat API running in Google Cloud Run, whichutilizes user chat history, local resources and outside resources. The application provides text summary of pdf and key points and communicates as a health care worker with users.
- Tracked and fixed bugs using Jira board, and sprints.
- handled scaling using Cloud run. The application uses python, openai and dspy.
- Wrote design document to communicate design and outcomes.
Python Machine Learning Developer
AlgoTech, L.L.C-FZ
- Contributed to the required enhancements and updates to use the latest packages and tools—the code was old, and the client had old software to do regression for Bitcoin pricing.
- Used parallel computing instead of threading as Python does not use threading well.
- Optimized the trained models with metrics to enhance performance using the Bayesian optimization method.
- Made the code faster and more robust as it used to crash each day.
Senior Python Developer
Johanson Dielectrics Incorporated
- Developed an application that optimizes the layout of capacitors in a Gerber file, meeting copper area and distance between polygons.
- Wrote the code in Python, using Numba for speed. Used shape transformation to optimize location.
- Generalized the code to work with any shape and given conditions.
- Wrote DXF file for the layout that meets area and distance conditions.
Machine Learning Engineer
NorthEast Monitoring Inc
- Developed MATLAB code to classify ECG signals using deep neural networks of mixed type that takes wave forms and other lab data.
- Generated results and selected models based on the F1 score.
- Built code that handles class imbalance for four classification types.
AI/ML Developer
Kevin O'Hagan
- Built an application to load emails from Outlook and create a way to extract information from emails. The information includes an estimate of the effort, a summary of the case, and a title.
- Extracted information from emails using large language models.
- Estimated the effort done in the case to calculate the cost.
NLP Expert
MeetMonk Inc
- Fine-tuned whisper voice to text to recognize the Indian language using medium and large models with Parameter Efficient Fine Tuning.
- Utilized Bark, XTTS, METAVOICE, and text-to-speech for the voice cloning of an Indian speaker to speak the English translation.
- Enhanced the MetaVoice-1B Facebook model with custom voice.
AI Expert
Getech Education Ltd
- Architected a note-taking and summarization solution using OpenAI and LangChain.
- Solved problems with summarizing large text from different documents.
- Architected the AWS solution using Lambda/Fargate ECS.
Machine Learning Developer
Gutify Ltd
- Used images with a multilabel associated with them, created a multilabel image classifier to classify images and obtain related tags.
- Deployed the model in the Google Cloud platform, with a schedule to run daily.
- Stored the images in Airtable so the solution obtains the new images from Airtable and labels them.
AI/Machine Learning Developer
Paragon Component Systems LLC
- Optimized the truss webs using Bayesian optimization, having an unknown number of webs, and optimized the cost maintaining that the structure passed strength and structure requirements.
- Worked with different truss structures and found an optimum number of webs to lower cost.
- Designed a system to find a matching of previously designed truss structure with the structure that passed strength requirements.
Senior Data Scientist
Alcoa
- Designed and built Alcoa Azure data science pipelines using TSFRESH for feature engineering, Boruta for feature selection, TPOT/Optuna for model selection, and Shap for feature importance.
- Attended stakeholder meetings, gathered requirements in A3 and DI, and created project objectives and deliverables.
- Peer-reviewed the team code and checked if the code achieved the desired deliverable.
- Built models and deployed them on-premises or Azure Cloud, sending output to PHD or OSI/PI (mining) to be consumed by dashboards for stakeholders.
- Built deep learning models for image processing from drones to detect bauxite quality and estimate sand and reactive silica using images.
- Preventively maintained, forecasted equipment failure, useful life, reliability, and optimization.
- Optimized mining with mixed integer linear programming (MILP) using CPLEX, Optuna AnyLogic, and spatial data analysis using ArcGIS Server/Esri.
- Worked on ETL/ELT using Azure Data Factory and FME for spatial data.
- Worked in three business verticals, mining, refining, and smelting.
- Used PyTorch, Pandas, NumPy, scikit-learn, TPOT, TSFRESH, Azure ML, and Azure Data Lake (Gen1 and Gen2).
Machine Learning Expert for Data Science POC
Breadcrumb Data Limited
- Developed time series anomaly detection using autoencoders for bearing and mills.
- Used a Controller Area Network (CAN bus) to build a vibration anomaly detection. The system uses the accelerometer acceleration data to estimate velocity and displacement.
- Estimated power spectrum to estimate the harmonics of vibration and use that to estimate the health of the equipment.
- Used TensorFlow to build this application and deployed it to Pi Zero and Pi 4.
Data Scientist and AI Consultant
S Wave International Corp
- Developed deep learning models to auto-tag music records using generated spectrogram images created from music records.
- Tracked and fixed bugs using Jira as a reporting tool.
- Deployed the production model in AWS Fargate with an elastic application load balancer to auto-scale the auto-tagging of music records depending on demand.
- Estimated the beat per minute using signal processing and deep learning. The project used the ViT model to classify spectrograms.
- Used PyTorch Lightning and DeepSpeed to train models using a large number of tracks.
GIS Platform Architect/Engineer
Birch Infrastructure, PBLLC
- Replicated data from Velocity Suite to BigQuery, performed many ETL operations, and generated materialized views using DBT.
- Converted geospatial data from shapefiles into BigQuery spatial.
- Generated a materialized view of LMP prices with weather information, updated hourly.
- Created Prefect flows to create jobs for data scrapping and downloading from various sources into BigQuery.
- Created APIs to perform data downloads from various sources like FRED, LMP prices, and other spatial data sources.
Developer
Toptal Client
- Created an API for a chatbot to train medical students with virtual patients for the Meksi project in Sydney.
- Used pre-trained BERT models with examples sent from the client.
- Created an API to infer the intent from user inquiries.
- Built a chatbot using the intent classifier, Named Entity Recognition (NER).
- Merged the Bert Response with an IBM Watson assistant response based on confidence to identify the intent. The results were 97% accurate across 512 intents.
Machine Learning Consultant
Oyu Tolgoi LLC
- Converted seismic traces into the spectrogram to be used with image detection, a Resnet50 model.
- Built a multi-input model with the height of the event as an embedding input and spectrogram as the second input. The model and approach will be published in the scientific journal Bulletin of the Seismological Society of America.
- Deployed the model successfully. It is planned to go into production in 2020.
Machine Learning Consultant
Global Unmanned System
- Developed algorithms to estimate above-ground biomass using point cloud data from drone images.
- Created object detection software for sealion detection in drone images.
- Classified images for sandalwood detection in drone images.
- Developed various image analysis software for drone images using OpenCV.
- Created various GIS applications for satellite images and point cloud data.
- Developed a program to help ship berthing at Fremantle Port using a SLAM algorithm and graph network.
- Gathered requirements and met with mine managers and refineries to learn about their problems and find possible projects.
- Converted raster layers into vector layers using kernel methods/OpenCV.
Senior Data Scientist
Freelance Work
- Developed refineries predictive maintenance using machine learning in Databricks, Azure Machine Learning service, and Azure Machine Learning Studio.
- Built time series prediction using Keras and PyTorch for anomaly detection.
- Built time series prediction with LSTM/CNN using multivariate one-minute sensors data.
- Developed a PowerBI dashboard for mining with Fleet Management System.
- Built a sound and vibration equipment health using a convolution neural network.
- Managed external contractors to evaluate cloud technology and perform a proof-of-concept solution to common anomaly detection in time series data and apply it to all pumps in the refinery.
Senior Data Scientist
Western Power
- Built energy demand time series prediction with multivariate half-hourly input data using LSTM/CNN neural networks.
- Created a production solution to use forecasted weather data to forecast demand deployed to AWS Fargate.
- Project-managed through Jira tickets and code stored in Bitbucket.
Senior Spatial Engineer
BHP
- Helped a big mining company to take advantage of its spatial data.
- Created driver behavior analysis software for a mining operation.
- Worked with natural language processing with Keras.
- Developed various GIS projects using ArcPy and C# ArcObjects.
- Created predictive models using machine learning and Apache Spark/Databricks.
- Completed a time series data analysis using Kalman filters for vehicle tracking.
- Mounted edge devices on diggers in an underground mine (Olympic Dam Mine) to classify the underground signs and determine if the bucket was full or empty.
- Converted raster layers of roads and vegetation into vector layers for processing in ArcGIS.
- Gathered requirements and met with mine managers and refineries to learn about their problems and find possible projects.
- Developed a Shiny (in R) dashboard for mining, a hotspot of high-rack events.
Senior Algorithm Engineer
Fugro
- Developed image analysis software for underwater object detection.
- Processed point cloud data using C++/Python for both underwater objects and above ground.
- Created image classification for a remote sensing Lidar point cloud using Python running in AWS and Fugro Roames for Ergon.
- Created C++ numerical algorithms for echo sounder calculation using Armadillo C++, OpenBLAS, and algorithms including Kalman filter, LAZ smoothing, and classification.
- Created various Julia and R regression machine learning applications.
- Converted raster layers underwater to detect cables.
Senior GIS Developer
Department of Mines and Petroleum
- Developed various GIS software to help surveyors in their work.
- Wrote classification and regression software for a GIS application.
- Developed GeoMap.WA, which is used to display the department's GIS products.
- Developed projects to do point cloud data analysis using ArcGIS software.
- Wrote various SQL server scripts to optimize retrieval of data.
Software Engineer
Western Power
- Supported GIS software to show Western Power Assets in Western Australia.
- Created predictive models for wooden pole maintenance and inspection using R.
- Developed and helped in the establishment of the wooden pole serviceability index.
- Wrote classification software for pole top fire prediction and pole serviceability index.
- Wrote Oracle scripts to download data for Oracle reports and optimize database queries.
Senior Software Engineer
Comsec
- Served as a senior software engineer and worked with various kinds of share trading software.
- Developed and designed NAB Equity Lending's online margin lending software.
- Supported various kinds of online trading software and managed funds.
- Created a SQL Server and Oracle database that shared procedures and database optimization.
- Participated in the design of NAB Equity Lending's migration to the Commsec Apollo project.
- Provided bug fixes and problem-solving for issues with trading software.
Senior Software Engineer
ERG
- Wrote Various C++ and Java applications for smart rider ticketing.
- Wrote transaction processing software in C++ to run under Solaris Unix and Windows operating systems.
- Supported existing software and bug fixes, code debugging, code executing speed (profiling), and peer review.
- Generated Oracle reports for Ventura bus travel times, loading, and peak time analysis.
- Optimized various PL and SQL queries for reports using Toad and Oracle query analysis.
Postdoctoral Research Fellow
Columbia University, New York
- Conducted research on the Columbia Linear Machine (CLM) as a postdoctoral research fellow at Columbia University.
- Supervised PhD and honor students and assisted with lectures and labs.
- Worked with LabVIEW on their National Instrument (NI) products control experiment.
- Looked after the lab supplies and placed orders to maintain operations.
- Wrote signal processing software (filtering) for probe measurements to remove noise.
- Wrote scientific papers and published the results of experiments.
Postdoctoral Research Fellow
Flinders University of South Australia
- Supervised PhD students and honour students during their studies.
- Wrote numerical analysis software for signal processing.
- Helped with lectures and lab tutorials and demonstrations.
- Maintained the lab by ordering supplies and repairs.
- Wrote scientific papers and published results of experiments.
Experience
Automatic Processing of Mine Micro-seismic: Application to the Oyu Tolgoi Mine
Our processing scheme mixes novel approaches and a more traditional method relying on energy variation-based event detection and on the picking of arrival times. Our platform directly employs the raw events detected by the seismic system. The event detection is set to be highly sensitive. When an event is detected, a block of continuous data is extracted and the event is located using a grid-based event location method. It classifies the event into ten categories, picks the arrival times, locates the events, and calculates the source parameters. The Oyu Tolgoi mine is currently using this system to process data in a production setting.
The system is deployed on Kubernetes clusters. Automatically processing seismic data requires the adoption of a holistic approach that extends to the design of the seismic array, the choice of sensors and the type and shape of the input data.
Using convolution neural networks and mixed networks with other features available, an automatic seismic identifier was created that outperforms other models in the market.
Auto-tagging Music Records
Electricity Demand Prediction
https://github.com/cobleg/Hack-A-GigWestern Power had a hack-a-gig competition to help them use their smart meter collected data, weather conditions, and population categories to determine their electricity demand. This helps clients to achieve minimum costs and helps Western Power plan its electricity network.
Mine was the winning solution and was written in Python using Keras for neural network and a mixed input model neural network, which consisted of the following inputs:
• Embedding of categorical fields like household categories, days to public holidays, and days after a public holiday (up to 5 days).
• Tabular numerical data for weather conditions.
• LSTM of the previous usage pattern of electricity.
These inputs were then concatenated into a fully connected deep neural network with one output variable to estimate the current usage of electricity per household. The data is then aggregated to reduce error by localities. My solution gave the best result in the competition, and the best reward was getting contracted by Western Power to do the project.
Driver Behavior Analysis
I wrote a C# application to collect GPS location data from various sources streamed through Kafka and then calculate the speed and orientation using the Kalman filter to reduce errors in the GPS measurements, and I stored the data in an Elasticsearch spatiotemporal database. I then used this data to determine events using the Elasticsearch database. I was the lead developer in the project.
Education
Master's Degree in Actuarial Science
Curtin University - Perth, Western Australia
Graduate Diploma in GIS
Curtin University - Perth, Western Australia
PhD in Science and Engineering
Flinders University - Adelaide, South Australia
Master's Degree in Communications Engineering
Cairo University - Cairo, Egypt
Bachelor's Degree in Communications and Electronics
Cairo University - Cairo, Egypt
Certifications
GCP Associate Cloud Engineer
GCP
Artificial Intelligence Nanodegree
Udacity
Skills
Libraries/APIs
Scikit-learn, PyTorch, Keras, TensorFlow, ArcGIS, XGBoost, Pandas, NumPy, Scikit-optimize, SpaCy, OpenCV, DeepSpeed, PyTorch Lightning, Natural Language Toolkit (NLTK), Dlib, Fast.ai, TensorFlow Deep Learning Library (TFLearn), Armadillo
Tools
GIS, Esri, Azure Machine Learning, MATLAB, Amazon SageMaker, Google Kubernetes Engine (GKE), Plotly, Google AI Platform, Amazon Elastic Container Service (ECS), FME, ArcGIS Runtime SDK for .NET, Microsoft Power BI, Spotfire, Kibana, Azure ML Studio, Cloudera, ChatGPT, BigQuery, Google Analytics, Bloomberg, Git, LabWindows/CVI, Protégé, Prefect, DataViz, CPLEX, Microsoft Exchange, MATLAB Statistics & Machine Learning Toolbox
Languages
C++, C#, R, Python, SQL, Python 3, JavaScript, Bash, T-SQL (Transact-SQL), Java, Fortran, SPARQL, OWL2, Arc
Frameworks
.NET, Optuna, Selenium, Hadoop, RStudio Shiny, Spark, Qt, DSPy
Paradigms
Agile Software Development, Anomaly Detection, ETL, Functional Programming, Radio Frequency (RF) Protocol, Agile
Platforms
Ubuntu, Azure, Jupyter Notebook, Oracle, Windows, Amazon Web Services (AWS), Google Cloud Platform (GCP), Linux, ArcGIS Server, Apache Kafka, Docker, NVIDIA CUDA, Kubernetes, Databricks, Hortonworks Data Platform (HDP), Raspberry Pi, AWS Lambda
Storage
Microsoft SQL Server, Elasticsearch, MySQL, PostgreSQL, Amazon S3 (AWS S3), Neo4j, PL/SQL, Redshift, Amazon DynamoDB, Redis
Industry Expertise
Project Management, Insurance
Other
BERT, Natural Language Processing (NLP), Artificial Intelligence (AI), Deep Learning, Statistical Learning, ArcGIS GeoEvent Server, ArcPy, Predictive Analytics, Data Science, Data Engineering, RStan, Image Processing, 3D Image Processing, RSelenium, Computer Vision, Machine Learning, Models, Generative Pre-trained Transformer 2 (GPT-2), Time Series Analysis, Data Analysis, Kalman Filtering, Sensor Fusion, Object Detection, Audio, Signal Processing, Video Processing, Forecasting, Statistics, Quantitative Finance, Finance, Data Analytics, Data Visualization, API Integration, Analytics, Physics Simulations, Simulations, Datasets, Language Models, Text Generation, Data Inference, Fine-tuning, Google BigQuery, Algorithms, Recommendation Systems, Communication, Open Source, Data Mining, Azure Databricks, Machine Learning Operations (MLOps), Data Reporting, Software Development, Leadership, GPU Computing, Architecture, AI Design, Time Series, Neural Networks, Artificial Neural Networks (ANN), Chatbots, Sentiment Analysis, Chatbot Conversation Design, Real-time Data, Data Processing Automation, Text Analytics, Web Scraping, Mathematics, Mathematical Analysis, Data Modeling, Financial Data, Word Embedding, Generative Pre-trained Transformers (GPT), Research, Geographic Information Systems, OpenAI GPT-3 API, Agile Product Delivery, Internet of Things (IoT), Hugging Face, Bayesian Inference & Modeling, Design Language, OpenAI, Technical Leadership, Regression Modeling, Large Language Models (LLMs), Text to Speech (TTS), LangChain, Uniformance Process History Database (PHD), OSI Model, Point Clouds, Reinforcement Learning, Generative Adversarial Networks (GANs), NAG Numerical Library, Graph Neural Networks, Generative Pre-trained Transformer 3 (GPT-3), Stable Diffusion, Causal Inference, Big Data, Real-time Vision Systems, Financial Modeling, Hardware, Computational Finance, Vehicle Routing, OpenAI GPT-4 API, Frameworks, Raster to Vector, Windows 10, GPS, Transformer Models, TAPAS, Data Build Tool (dbt), Variational Autoencoders, ArcSDE, AIOps, Data, Feature Engineering, IOTA, Autoencoders, CAN Bus, Text to Image, Signal Protocols, Electronics, Digital Communication, Vector Databases, Pinecone, Speech to Text, Speech Recognition, Microsoft Graph API, NLU, Deep Neural Networks (DNNs)
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring