AI Engineer with 2 years of experience building production AI systems across machine learning, Generative AI, RAG, and multi-agent architectures. Skilled in retrieval engineering, prompt engineering, LLM evaluation, Mode…
AI Engineer with 2 years of experience building production AI systems across machine learning, Generative AI, RAG, and multi-agent architectures. Skilled in retrieval engineering, prompt engineering, LLM evaluation, Model Monitoring, and MLOps, using LangChain, LangGraph, and transformer-based models deployed on AWS. Experienced in collaborating with QA, DevOps, and business teams across development, UAT, and post-deployment support.
<p>Built a production retrieval pipeline over 5,000+ compliance documents using Sentence Transformer embeddings indexed in FAISS and Pinecone, with hybrid search (BM25 + dense vectors) lifting Precision@K by 20% over a dense-only baseline.<br>
Integrated GPT-4o for grounded answer generation with source citations, sustaining p95 latency under 4 seconds at 500+ daily queries, with auto reindexing to maintain index freshness.<br>
Established MLflow experiment tracking across 100+ evaluation runs and retrieval benchmarking pipelines to monitor search relevance and model performance.</p>
<p>Built Random Forest classifier (85% accuracy) using 25+ engineered features (description completeness, severity alignment, reproducibility); boosted high-quality bug submissions by 30% via real-time feedback integration.<br>
Engineered XGBoost regression + constraint-based optimizer predicting optimal team sizes across 15+ pillars (500+ person project); integrated ARIMA forecasting for demand prediction, driving 10% revenue uplift.</p>
<p>Achieved a 60% reduction in manual defect triage time by architecting a reusable multi-agent framework with LangGraph, effectively lowering the process from 4 hours to under 90 minutes.<br>
Delivered 60+ features and 50+ backend fixes across development and QA environments, optimizing SQL performance and hardening enterprise API integrations to reduce recurring defects.<br>
Fine-tuned a transformer classifier with QLoRA on a 5,000+ record QA bug dataset, raising F1-score by 18% for valid vs. false-positive defect detection.</p>
نظام التوظيف التشغيلي المعتمد على الذكاء الاصطناعي أولاً، الذي يوحّد ويؤتمت دورة التوظيف بالكامل للمؤسسات الحديثة حول العالم.
د.جوب منطقة حرة ذ.م.م. 2026 © جميع الحقوق محفوظة
