Souha Kacem

Led the integration and orchestration layers that dynamically route between LLMs, reducing 40h manual workflows to 30min with 90% accuracy, for PwC, KPMG, ADX clients. Engineered multi-agent system with retrieval augmented prompting, boosting contextual reasoning and accuracy by 35% with Streamlined CI/CD pipelines. Developed an AI-driven auto-rating assessment system combining speech analysis, semantic analysis, and (facial expression detection, improving evaluation precision by 45%.

• Led data analytics and marketing intelligence projects for e-commerce clients using Python, SQL, and Power BI. • Delivered actionable insights that increased client revenue by 40% through customer segmentation and predictive analysis. • Designed interactive dashboards and reports, transforming raw data into clear, executive-ready insights that accelerated decision-making.

• Automated data ingestion pipelines and web scraping workflows, processing 20K+ lead profiles per cycle, feeding Power BI dashboards for real-time KPIs. • Applied clustering algorithms (K-Means, Agglomerative) to segment leads by digital maturity, optimizing marketing strategy targeting accuracy by 30%. • Partnered cross-functionally with business and marketing teams to align data initiatives with company goals, improving decision-making efficiency.

• Built a multilingual AI chatbot supporting 40+ languages, integrating IBM Watson Assistant & Translator — reducing customer response time by 60%. • Enhanced NLP performance through improved intent classification and dialogue management workflows.

Designed a facial liveness detection system using CNNs and Transfer Learning for KYC verification. Implemented advanced anti-fraud measures: thermal detection, 3D face analysis, and expression/movement recognition using OpenCV.