AI Engineer and Data Scientist specializing in Generative AI, LLM applications, agentic systems, RAG, and production-grade AI solutions.
I have hands-on experience designing and deploying end-to-end AI systems using Pyt…
AI Engineer and Data Scientist specializing in Generative AI, LLM applications, agentic systems, RAG, and production-grade AI solutions.
I have hands-on experience designing and deploying end-to-end AI systems using Python, FastAPI, LangGraph, OpenAI, Gemini, vector databases, OpenSearch, Docker, GCP, Redis, and modern LLM infrastructure. My work includes multi-agent systems, AI-powered automation, retrieval and reranking pipelines, autonomous decision systems, content-generation platforms, and conversational AI.
I combine a strong mathematical and machine-learning background with practical software engineering and product experience. I am comfortable taking AI projects from experimentation and evaluation through architecture, API development, deployment, monitoring, and production optimization.
Core expertise: Generative AI • LLMs • AI Agents • LangGraph • RAG • NLP • Machine Learning • Deep Learning • Python • FastAPI • OpenAI • Gemini • OpenSearch • Vector Databases • Docker • GCP • Redis • SQL • Data Engineering • AI System Design
<p>Technical counterpart on architecture, cross-entity products, and the vendor/stakeholder decisions that decide what actually gets deployed.<br>
Evaluated SambaNova and Cerebras on latency, throughput, and model fit as the AI technical counterpart.<br>
Advised directors on integration, including Azure OpenAI / Microsoft Copilot against real government use cases.<br>
Shipping a government-wide meeting-intelligence platform serving 20,000+ users: STT, speaker diarization, and a multi-stage pipeline for minutes, action items, insights, decisions, and follow-ups.<br>
Architected production multi-agent systems on Deep Agents / LangGraph with skills, tool calling, and MCP: agents that plan, invoke tools, and coordinate end-to-end government workflows.<br>
Self-hosted open-source models on-prem with vLLM for sovereign government workloads: controlled inference, latency/throughput tuning, and no vendor lock-in on the serving path.<br>
Promptfoo as a merge gate: golden evals, production traces, prompt refinement, guardrails, and prompt-injection red-teaming before AI products ship.<br>
Azure AI Search with custom enrichment for Global Search.</p>
Coding used: R
Sport analysis and classify the teams that are underperforming and outperforming
Predicting the final ranking of the teams by using the results until the mid-season
<p>Built the Autonomous Merchant into a market-first autonomous commerce stack with 3,000+ orders fulfilled end-to-end.<br>
Wired acquisition, creative, and media buying into one hands-off loop.<br>
Owned the AI roadmap.<br>
Led recruitment for all AI profiles (Junior to Senior) and mentored engineers on code quality, architecture patterns, and best practices.<br>
Architected a Retrieval-Augmented Generation (RAG) platform using LLMs and external corpora to synthesize domain-expert marketing agents.<br>
Developed multi-step planning workflows to generate optimal media-buying strategies.<br>
Built an end-to-end AI creative generator transforming structured product data into social-media assets (video, automated voiceover, subtitles) using ffmpeg and VLM models.</p>
<p>Co-founded Qanu.ai, an AI-powered legal workspace: playbook contract review and agentic document editing over Saudi statutory and case law.<br>
Shipped playbook-driven contract review: custom and imported rules, traffic-light findings, and clause-anchored evidence in the live editor so lawyers can run a negotiation playbook on real paper.<br>
In-editor agent proposes merge, split, and rewrite as tracked changes in the collaborative editor; lawyers accept or reject without leaving the document.<br>
Architected a bilingual Arabic NLP / English RAG pipeline for legal QA.<br>
Hybrid retrieval over statutory and case law using OpenSearch: embeddings, vector search (BM25 + kNN), RRF, and reranking.<br>
Implemented a streaming chat engine with LangChain, SSE, and an AI SDK-compatible protocol.<br>
Cross-lingual QA with source citation and highlight linking.<br>
Wired Langfuse for LLM observability, traces, and quality evaluation.<br>
Built ingestion pipelines over 65k+ judicial decisions and 2,300+ legal principles.<br>
Metadata schemas for court/jurisdiction filtering, and background indexing for user-uploaded documents.</p>
<p>GenAI integration and operational forecasting for industrial gases and healthcare.<br>
Architected and deployed a RAG-powered GenAI onboarding chatbot (LangChain/LangSmith), achieving high-precision retrieval and adoption by 500+ employees.<br>
Engineered and globally deployed an internal Python package for automated Google Sheets interaction, enhancing reporting efficiency across 50+ sales teams worldwide.<br>
Designed and implemented a high-volume IoT telemetry pipeline via REST APIs, generating €100,000 in annual cost savings by eliminating reliance on external data services.<br>
Deployed predictive models for customer bulk orders to improve demand planning and drove adoption of MLOps and clean code standards, mentoring junior data scientists.</p>
<p>Forecasted DOOH underspend in Dataiku, uncovering €1,000,000/year leakage.<br>
Async Python/Pydantic ingestion of 1 million+ campaign assets.</p>
<p>Create a detection intrusion system<br>
Simulate a network of virtual machines to simulate attacks such as DDos and ARP poisoning in Google Cloud Plateform<br>
Create a dataset by collecting traffic in real-time with Kafka Streams<br>
Labelize a dataset with Snorkel and use machine learning algorithms to detect attacks<br>
Intrusion detection system with machine learning algorithms</p>
<p>Built a Kafka Streams intrusion-detection pipeline (Voting Classifier) on a GCP threat testbed (DDoS, ARP poisoning), with Neo4j dashboards for topology and live alerts.</p>
<p>Coding : Python<br>
Manage a big dataset (more than 4 millions of observations and 9 variables)<br>
Create new aggregated variables<br>
Used many machine learning algorithms such as XGBoost, Logistic Regression, MLPClassifier, Extra Trees Classifier, SVC to classify the days when it's raining and then use quantile regression to determine the rainfall</p>
<p>Coding used : Python<br>
We used time series model such as ARIMA and LSTM neural network to predict the amount of future sales of a company</p>
Create a detection intrusion system
Simulate a network of virtual machines to simulate attacks such as DDos, ARP poisoning in Google Cloud Platform
Create a dataset by collecting traffic in real-time with Kafka Streams
Labelize a dataset with Snorkel and use machine learning algorithms to detect attacks
Intrusion detection system with machine learning algorithms
Coding: Python
Manage a big dataset (more than 4 million of observations and 9 variables)
Create new aggregated variables
Used many machine learning algorithms such as XGBoost, Logistic Regression, MLP Classifier, Extra Trees Classifier, SVC to classify the days when it's raining and then use quantile regression to determine the rainfall
Coding used: Python-TensorFlow-Keras-Google Colab
Instantiation of a deep neural networks by using a GAN model
Creation of multiples fakes faces
The AI-first Recruitment Operating System that unifies and automates the entire hiring lifecycle for modern enterprises worldwide.
Dr Job FZ LLC. 2026 © All Rights Reserved
