I design and ship agentic AI systems end to end: multi-agent pipelines, retrieval-augmented generation, and self-hosted inference infrastructure. My open-source systems run entirely on local models, with no external API dependency and no per-call vendor cost.
Currently architecting LLM-powered automation at Capgemini. M.Eng. in AI and Data Science, University of Ottawa. AWS Certified Machine Learning Specialty.
Cairo, Egypt · Open to AI architecture roles · Arabic & English
Growth Engine · nine-agent outbound automation
An autonomous B2B prospecting system. Nine specialised agents discover leads, research them, write personalised outreach with a local LLM, send it under hard-coded deliverability rules, then watch for replies and report on the funnel.
PROSPECTOR → ENRICHER → VERIFIER → RESEARCHER → WRITER
→ SENDER → REPLY WATCHER → REPORTER → DASHBOARD
Every stage is an independently testable unit passing state through a shared persistence layer. The LLM layer is provider-agnostic with per-role model assignment, defaulting to local Ollama and extending to OpenAI, Anthropic, Groq or any OpenAI-compatible endpoint, with daily caps and automatic local fallback. Compliance is architectural, not bolted on: warm-up ramps, randomised send jitter, restricted send windows, suppression lists, and a circuit breaker that halts a campaign above a 3% bounce rate. Execution safety is three-tier, with two-key confirmation before anything goes live.
FastAPI APScheduler SQLAlchemy 2.0 Alembic Ollama aiogram 3 PostgreSQL Pytest
YouTube Shorts Studio · self-hosted multimodal video pipeline
Turns one long-form video into multiple publish-ready vertical Shorts, on hardware you already own.
INPUT → TRANSCRIBE → SEGMENT → MONTAGE → REFRAME
→ CAPTION → OVERLAY → METADATA → THUMBNAIL → UPLOAD
The interesting constraint was memory. Whisper, the LLM and the background-removal model all want the GPU at once, so scheduling them through a shared VRAM budget lets the whole stack run on a single 8 GB consumer card. The segmentation LLM returns sentence indices rather than timestamps, which guarantees cuts land on semantic boundaries and removes the dominant failure mode of timestamp-based approaches. Captions render right-to-left Arabic with word-level highlighting synced to Whisper word timestamps.
Whisper large-v3 Ollama Command-R Qwen 2.5 FastAPI OpenCV FFmpeg rembg yt-dlp
A Dockerised assistant handling text conversation, image captioning, image transformation and text-to-image generation entirely on local open-source models. n8n orchestrates the workflow and PostgreSQL persists conversational memory, so context and personalisation survive across sessions. Nothing leaves the machine, which makes it private by construction and free per message.
LLaMA Gemma ComfyUI Flux n8n PostgreSQL Docker
Generative AI and LLMs
Agentic systems and orchestration
Infrastructure and MLOps
Data and analytics
Thirteen repositories from my M.Eng. and undergraduate work, each pairing the source code with the reports, papers and presentations that document it.
| Repository | What it covers |
|---|---|
| AI-For-Cybersecurity | CT-GAN intrusion detection, Kafka model streaming, DCGAN anomaly detection |
| Language-Identification-System | Seven-language speech ID: dataset pipeline, MFCC/Mel modelling, Flask deployment |
| Smart-Cities-Simulation | Crowdsensing pipeline: task generation, CGAN synthesis, anomaly detection, ns-3 |
| NLP-Projects | Authorship attribution, book clustering, ontology-driven chatbot |
| Big-Data-Spark | Distributed processing in Scala, PySpark and Spark SQL with streaming jobs |
| Computer-Vision | Transfer-learning classification plus KNN and filters built from scratch |
| Machine-Learning-From-Scratch | Decision trees, boosting, KNN and OVR/OVO from first principles |
| Data-Science-Analytics | Sentiment-driven price prediction, churn analysis in R, SQL analytics |
| Deep-Learning | Neural architectures on the UNR-IDD intrusion detection dataset |
| Azure-Synapse-Streaming | Cloud warehousing and real-time event processing on Azure |
| Reinforcement-Learning-Projects | Agent trained to drive through environment interaction |
| Social-Media-Scraper | Batch-downloads 100+ videos per run with automatic trimming |
| Mongodb-Cassandra | Document versus wide-column NoSQL data modelling |
AI and Automation Engineer · Capgemini · 09/2025 to Present
Architect and deploy end-to-end AI automation systems connecting multiple business applications, replacing manual processes with orchestrated, monitored pipelines. Design intelligent chatbots and voice assistants for client-facing platforms, owning conversational flow, state handling and interaction models. Integrate LLMs into chatbot architectures to improve contextual understanding and grounding across multi-turn conversations. Maintain development, staging and production environments, and author the technical documentation and workflow architecture diagrams that enable handover and reuse.
Data Analyst · Capgemini · 06/2023 to 09/2025
Automated recurring manual Excel reporting with Python, reducing a six-hour process to seconds. Developed SQL database solutions powering real-time Power BI dashboards, and analysed heterogeneous data sources to surface the patterns, trends and anomalies behind data-driven business strategy.
M.Eng. Artificial Intelligence and Data Science · University of Ottawa, Canada · 02/2022 to 02/2023 · Grade: Excellent
B.Sc. Computer Science · Misr University for Science and Technology · 09/2017 to 07/2021 · GPA 3.80 / 4.00
- AWS Certified Machine Learning Specialty · Amazon Web Services, May 2022
- AWS Certified Cloud Practitioner · Amazon Web Services, February 2022
- Microsoft Power BI · October 2023
- Leadership and Management · Dale Carnegie, May 2022
