Akram Assaf

Applied AI & Data Science to directly support and enhance Bayt's recruiting platform, bridging theory with production systems. Recruiting Intelligence & Behavioral Analytics: • Built large-scale behavioral models of job seeker and employer interactions • Developed temporal engagement analysis, funnel modeling (search → view → apply → interaction) • Identified key dynamics: timing-driven visibility, CV completeness and freshness impact • Translated insights into ranking strategies, product interventions, and monetization-aware features • Techniques: Cohort segmentation, temporal modeling, funnel analytics, feature engineering for ranking systems, exploration vs exploitation strategies NLP – Skill & Knowledge Extraction: • Developed a production-relevant NLP system for extracting structured skills from unstructured text • Architecture: Transformer encoder (BERT-based), Multi-task learning (MTL) with separate task heads, BiLSTM layers + CRF decoding for structured prediction • Techniques: span detection (BIO/BILOU tagging), handling overlapping entities and class imbalance • Optimization: dynamic loss weighting, learning rate scheduling, full fine-tuning of PLMs • Outcome: Improved accuracy of skill extraction, enabled downstream applications in matching and career intelligence Graph Algorithms & Network Modeling: • Applied graph theory to solve optimization and structural problems • Built graph models for route optimization, demand distribution, node importance analysis • Techniques: shortest path algorithms, geodesic calculations, centrality metrics (betweenness, degree, harmonic), clustering and community detection • Architecture: Neo4j for graph processing, MongoDB for flexible data storage, Redis for caching and real-time performance Data Systems & Architecture: • Designed polyglot data architectures combining graph databases, NoSQL systems, and in-memory caching layers • Focused on real-time responsiveness, scalability, and system reliability Research Design & Applied Data Science: • Developed structured methodologies for problem framing, hypothesis definition, and experimental design • Emphasized business impact, measurable outcomes, and cost vs value trade-offs • Broader exposure: Predictive modeling and statistical analysis, data visualization and storytelling, cross-domain applications (logistics, sustainability, consumer behavior), evaluation frameworks for ML systems

Coursework included software engineering, algorithms, data structures, and computer architecture. Developed a capstone project on distributed systems, showcasing skills in collaborative software development and problem-solving.