Khalid H.

Context: An international digital platform was struggling with LLMs generating inaccurate, inconsistent responses and insufficient user personalization across multilingual, multi-format content (text, audio, video, and images).

Actions: Analyzed user behavior, search journeys, and query intent to identify UX friction points. Fine-tuned LLMs and optimized prompts to improve accuracy and contextual relevance. Built a quantitative and qualitative artificial intelligence (AI) audit framework (fact-checking, reliability scoring, hallucination reduction). Ran A/B tests and longitudinal analyses to measure product improvements. Built Power BI KPI dashboards to track business impact.

Results: Measurable reduction in LLM hallucinations. Improved content relevance confirmed by satisfaction metrics. A reusable AI audit framework adopted by international product teams.

Stack: Python, SQL, LLMs, Fine-tuning, Azure, Power BI, A/B Testing, Machine learning, Deep Learning, MLOps, model deployment, prompt engineering, NLP

Context: An IoT startup specializing in battery innovation had no centralized view of its multi-source sensor data. Technical and business teams lacked visibility on operational performance, making maintenance reactive and costly.

Actions: Designed and automated robust ETL pipelines on Azure/Databricks integrating IoT sensors and external APIs in near real-time. Refactored existing architectures to reduce processing times. Developed Python/SQL scripts for automated KPI calculation and anomaly detection. Built interactive Power BI dashboards (DAX, Power Query) for performance monitoring and decision support.

Results: Full centralization of multi-source data. Reduced pipeline processing times. Proactive anomaly detection on battery lifecycles enabling better maintenance planning. Dashboards are fully adopted by business and technical teams.

Stack: Python, SQL, Azure, Databricks, ETL, Anomaly Detection, Power BI, DAX, Machine learning, Artificial Intelligence, predictive maintenance, statistical analysis, data visualization

Context: A retail client needed to optimize its store location strategy and monitor commercial performance. Data was scattered across internal sources, geographic datasets, and external APIs with no consolidated view for strategic decisions.

Actions: Built ETL pipelines using Python, SQL, and BigQuery on GCP, integrating multi-source and geomarketing data via APIs. Developed predictive ML models to identify high-potential commercial zones and anticipate sales trends. Conducted FinOps analyses to optimize costs and budgets. Created Power BI dashboards delivering predictive insights on sector performance and optimal location strategies.

Results: High-potential store locations were identified through predictive models. Improved cost management through FinOps analysis. Data-driven culture reinforced across operational teams trained on dashboards.

Stack: Python, SQL, GCP, BigQuery, ML, Geomarketing, FinOps, Power BI, Machine Learning, Artificial Intelligence, Deep Learning, business intelligence, geospatial analysis, reporting