Ahmed is available for hire
Hire AhmedAhmed Shaher
Verified Expert in Engineering
Data Scientist and Machine Learning Developer
Riyadh, Riyadh Province, Saudi Arabia
Toptal member since June 8, 2021
Bio
Ahmed is an applied AI leader with 10 years of experience building production AI systems at Microsoft scale and delivering enterprise AI platforms across the Gulf. He leads the OSOS AI Platform at Mozn, owning multi-GPU LLM inference, retrieval, and agentic orchestration deployed into air-gapped government environments. He works with clients from technical discovery through production rollout, and holds a master's in machine learning from Georgia Tech.
Portfolio
Mozn
Python, PyTorch, vLLM, LLM Agents, AI Agents, Agentic AI, RAG Systems...
Microsoft
Data Science, Machine Learning, Azure Machine Learning, Deep Learning, Python...
Microsoft
Python, PyTorch, Machine Learning, Deep Learning...
Experience
- Python - 10 years
- Machine Learning - 9 years
- Data Science - 9 years
- Deep Learning - 7 years
- Machine Learning Operations (MLOps) - 5 years
- Large Language Models (LLMs) - 3 years
- Agentic AI - 2 years
- Large Language Model Operations (LLMOps) - 2 years
Work Experience
AI Platform Lead
2025 - PRESENT
Mozn
- Led the OSOS AI Platform team, owning the architecture behind every AI capability on the platform: LLM and vision-language inference on multi-GPU A100 clusters, bilingual embedding serving, and hybrid vector and keyword retrieval.
- Designed the document processing and orchestration layer that decomposes analyst tasks into workflow trees, runs ReAct agents at the nodes, and reduces results into Arabic executive summaries and committee reports.
- Deployed into air-gapped, on-premises, and cloud environments for Saudi government and enterprise clients, and act as the technical escalation point for production incidents inside customer estates.
- Cut model and agent deployment time from three weeks to under one hour by standardizing the path from experiment to production release.
- Led customer-facing technical discovery, proofs-of-concept, and executive demonstrations across a pipeline of 10+ enterprise and government engagements.
Technologies: Python, PyTorch, vLLM, LLM Agents, AI Agents, Agentic AI, RAG Systems, Retrieval-augmented Generation (RAG), Vector Databases, Kubernetes, Docker, PostgreSQL, Redis, Machine Learning Operations (MLOps), Natural Language Processing (NLP), Optical Character Recognition (OCR), Text-to-SQL, Solution Architecture, GPU Computing, Artificial Intelligence (AI), Large Language Model Operations (LLMOps)
Applied Science Manager
2024 - 2025
Microsoft
- Owned the science roadmap for Microsoft's ranking infrastructure serving 100 million ranking queries per day, delivering an 8% improvement in availability SLA across engineering and science teams.
- Led a team building ML models that extracted behavioral interest signals from app usage patterns, improving retention across five Microsoft consumer applications serving 800,000+ daily active users globally.
- Managed and grew a team of 5 applied scientists and machine learning engineers, owning hiring, performance reviews, and technical mentorship while staying hands-on in model design and code review.
Technologies: Data Science, Machine Learning, Azure Machine Learning, Deep Learning, Python, Recommendation Systems, Classification, Clustering, Azure, Computer Engineering, Data Analysis, Machine Learning Operations (MLOps), Natural Language Processing (NLP), Data Visualization, Algorithms, SQL, Reinforcement Learning, Software Development, Design Patterns, Software Deployment, Azure SQL Databases, Data Modeling, PyTorch, Predictive Analytics, Artificial Intelligence (AI), Large Language Models (LLMs)
Senior Applied Scientist
2018 - 2024
Microsoft
- Drove a 20% increase in daily active users and a 1% lift in click-through rate by improving personalization and quality of news notifications, with a 15% overall gain in notification quality.
- Built early enterprise LLM proofs-of-concept for Microsoft's strategic customers that demonstrated capability and secured executive buy-in.
- Improved NER benchmark F1 by 4% through state-of-the-art research integration, and built an experimentation framework that reduced hyperparameter search space tenfold.
- Co-developed applied AI solutions with Microsoft's sales innovation team, including SMB credit risk assessment from transactional data and meta-learning for live-stream video highlight detection.
Technologies: Python, PyTorch, Machine Learning, Deep Learning, Natural Language Processing (NLP), Large Language Models (LLMs), Named-entity Recognition (NER), Recommendation Systems, Azure Machine Learning, Azure, Machine Learning Operations (MLOps), Personalization, Artificial Intelligence (AI)
Data Scientist
2018 - 2018
Elves
- Developed custom named entity recognition for airports and airlines to annotate travel queries and automate the flight booking flow.
- Built a reporting system for exploratory and predictive analytics to find trends and anomalies in the data. The system also reported on churning users and identified common scenarios to re-engage with a segment of churning users.
- Trained an LSTM spam classifier to identify users who have malicious behavior.
Technologies: Python, Data Analysis, Machine Learning, Classification, Clustering, Computer Engineering, Named-entity Recognition (NER), Deep Learning, Algorithms, Reinforcement Learning, Software Development, Data Science, Software Deployment, SQL, NoSQL, Plotly, Data Modeling, Predictive Analytics, Artificial Intelligence (AI)
Data Scientist
2017 - 2018
Benchmark Middle East
- Developed a book consultant chatbot (sipof.ink) that won first place in the Facebook Middle East and Africa Bots for Messenger Challenge in Productivity and Utility.
- Trained a Doc2Vec model on book data to find similarities and similar books to recommend.
- Trained an end-to-end memory network (MemN2N) for book recommendations and book question answering.
Technologies: Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Algorithms, Python, Machine Learning, Software Development, Computer Engineering, Data Science, Software Deployment, Classification, SQL, NoSQL, Data Modeling, Predictive Analytics, Artificial Intelligence (AI)
Junior Data Scientist
2016 - 2017
Cognitive
- Created a product linkage pipeline consisting of three stages—indexing, matching, and classification—to enable the creation of a master product from different eCommerce websites.
- Achieved 77% accuracy on products in both Arabic and English across three different eCommerce websites.
- Enabled users to compare products on different eCommerce websites, using this three-stage pipeline to match products with as little data as title and price range and create the master product page.
Technologies: Python, Machine Learning, Clustering, Classification, Text Classification, Algorithms, Software Development, Computer Engineering, Data Science, Software Deployment, Data Modeling, Predictive Analytics, Artificial Intelligence (AI)
Project History
Credit Risk Assessment and Limit Prediction for Fintech
An MLOps pipeline I created, including a machine learning model that used small and medium business transactions, assessed their credit risk classifications, and determined the proper limits for payment extensions. This increased access to short-term loans for the fintech by 24% and decreased the default rate by 5%.
Agentic Arabic Document Analysis for Government Clients
An orchestration layer that decomposes an analyst's request into a hierarchical workflow tree, executes React agents at each node against retrieved source documents, and reduces the results upward into an executive summary or committee memo in Arabic. It's built on durable parent-child workflows with a registry-driven node factory, deployed into air-gapped government environments with no external network access.
Multi-GPU LLM Inference Platform
The serving stack behind an enterprise AI platform: LLM and vision-language model inference across A100 clusters using tensor parallelism and FP8 quantization, bilingual embedding serving, and hybrid vector and keyword retrieval. Capacity and concurrency tuned against customer latency and throughput targets, running in on-premises and air-gapped deployments alongside cloud.
Dashboard for User Purchasing and Churn Analyses
Developed a reporting system for exploratory and predictive analytics to find trends and anomalies in the data. The system also reported churning users and identified common scenarios to re-engage with a segment of churning users based on their previous purchasing patterns, accounting for seasonality. The dashboard was created in Python and visualized on Google Data Studio. Using this project, the marketing team was able to promptly engage with users and decrease the purchase churning percentage by 24%.
Textual Emotion Recognition Using Ensemble Classifier
Implemented an ensemble classifier system to classify whether the emotion in an input text is anger, disgust, fear, sadness, or joy. The system is based on three learners: two statistical learners and a knowledge-base learner, ensembled in a majority voting approach. The ensemble was able to classify emotions from tweets with a 95% F1 score.
Education
2019 - 2021
Master's Degree in Computer Science
Georgia Institute of Technology - Atlanta, Georgia, USA
2011 - 2016
Bachelor's Degree in Computer Engineering
Ain Shams University - Cairo, Egypt
Certifications
JANUARY 2021 - JANUARY 2023
Microsoft Azure Developer Associate
Microsoft
MARCH 2020 - MARCH 2022
Microsoft Certified: Azure Data Scientist Associate
Microsoft
Skills
Libraries/APIs
PyTorch, Natural Language Toolkit (NLTK), vLLM
Tools
Azure Machine Learning, Plotly, Named-entity Recognition (NER), Text-to-SQL
Languages
Python, SQL
Platforms
Azure, Kubernetes, Docker, Linux
Paradigms
Design Patterns
Storage
Azure SQL Databases, NoSQL, PostgreSQL, Redis
Other
Machine Learning, Reinforcement Learning, Deep Learning, Software Development, Computer Engineering, Data Science, Software Deployment, Data Analysis, Natural Language Processing (NLP), Clustering, Classification, Text Classification, Ensemble Methods, Recommendation Systems, Algorithms, Data Modeling, Predictive Analytics, Artificial Intelligence (AI), Generative Pre-trained Transformers (GPT), AI Agents, Agentic AI, Large Language Models (LLMs), Machine Learning Operations (MLOps), Financial Data, GraphDB, Data Visualization, Time Series Analysis, Large Language Model Operations (LLMOps), Trend Analysis, Optical Character Recognition (OCR), LLM Agents, RAG Systems, Retrieval-augmented Generation (RAG), Vector Databases, Solution Architecture, GPU Computing, Personalization
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