Mariam EL ALAOUI

• Redesigned the company’s data labeling and classification pipeline using Python-based ETL processes, reducing manual processing time by 98% and saving over 50 hours per 1,000 metrics. • Architected a modular Model Context Protocol (MCP) ecosystem that supports multi-agent AI workflows for automated root-cause analysis and telemetry-driven decision support. • Developed and deployed an agentic Retrieval-Augmented Generation (RAG) pipeline using ReAct reasoning, pgvector retrieval, and API integration, achieving 80% top-5 recall and reducing incident investigation time by 35–40%. • Built a scalable pgvector-based retrieval system integrated with AWS Bedrock to enhance context-aware troubleshooting for engineering, DevOps, and observability teams. • Optimized cloud infrastructure through concurrency scaling, batching, hybrid CPU/GPU scheduling, and PySpark-based distributed computing, increasing system throughput from 1 metric/minute to 20–25 metrics/minute. • Generated over 100,000 synthetic telemetry and anomaly-detection samples to validate machine learning models, improving anomalydetection performance and reducing manual service-discovery tasks by 90%. • Delivered enterprise-level recommendations on AI governance, data modeling, and cloud-cost optimization, decreasing storage and token consumption expenses by 60% across multiple projects. • Collaborated with C-suite executives, product engineering teams, and infrastructure teams in Agile practices to align ML architecture, deployment pipelines, business continuity, and system-performance objectives. • Built internal dashboards, documentation, and technical playbooks integrated with cloud data warehousing solutions to improve system adoption, cross-team visibility, and repeatable engineering workflows.

• Conducted gap assessment between current analytics workflows and target performance outcomes, enabling the R&D division to prioritize AI initiatives aligned with revenue and efficiency goals. • Built predictive models for risk forecasting and process optimization, producing insights that informed financial planning and resource allocation strategies. • Authored internal research papers that shaped the company’s AI governance and innovation roadmap, aligning data investments with corporate digital-transformation objectives. • Partnered with business units to translate model outputs into executive dashboards supporting KPI monitoring and strategic decisionmaking. • Analyzed IoT telemetry and network sensor data to design streaming applications for predictive maintenance and anomaly detection across distributed infrastructure systems.

– Developed and optimized real-time, low-latency AI agents using LLMs and multi-agent frameworks (LangChain, CrewAI). – Fine-tuned multilingual NLP models for adaptive, context-aware interactions. – Built lightweight voice cloning and persona modeling systems with minimal data. – Collaborated on agent-native security models and scalable orchestration pipelines (MLOps, Docker, AWS). – Contributed to auto-refinement loops and benchmarking against platforms like OpenAI and Google. – Contributed to internal white papers and knowledge-sharing efforts, aligning research with long-term innovation strategy. – Utilized Python, SQL, TensorFlow, PyTorch, ServiceNow, Snowflake, Jira, Helix, AWS in hybrid cloud environments.

– Developed a healthcare Android chatbot using LangChain, Pinecone, GPT-3.5 Turbo, and Llama 2. – Processed data from PDFs, created embeddings with Sentence-Transformers, and stored them in Pinecone. – Improved query precision by 20% and reduced processing time by 20%, achieving an average response time of 7.44 seconds. Enhanced user satisfaction with an 8/10 average score. – Managed data-driven projects with a focus on accuracy and efficiency, ensuring high-quality results in fast-paced environments. Utilized Python, Pytorch, NLP, Transformers, Hugging Face, and Google Cloud for automatic translation.

Sentiment Analysis on Breast Cancer: – Conducted sentiment analysis on breast cancer-related tweets using Python, Scikit-learn, and NLTK. Scraped data with Twitter API and Tweepy, handling it with Pandas. – Employed machine learning models (XGBoost achieving 79% accuracy, SVM) and visualized insights with Matplotlib. Managed project in Jupyter Notebook and version control with Git

– Analyzed production efficiency and resource use to optimize processes.Cleaned large and complex datasets using Python (Pandas, NumPy) to identify bottlenecks. – Configured and maintained proprietary software and hardware systems, ensuring smooth installation and integration of technologies. Presented insights through Power BI dashboards for data-driven decisions