Mohamad Bilal

Developed an AI-powered legal document analysis system that automatically identifies risks and potential issues in legal documents using advanced machine learning and natural language processing techniques. Architected and implemented a comprehensive knowledge base section utilizing vector embeddings for efficient semantic search, enabling accurate retrieval of the most relevant legal precedents and documentation. Integrated Langfuse for comprehensive LLM observability and tracking, monitoring model performance, token usage, latency metrics, and user interactions to optimize system accuracy and efficiency. Built intelligent retrieval mechanisms that match legal documents with the best-fit precedents and case studies from the knowledge base, significantly improving legal analysis workflow and reducing manual review time.

Increase prediction accuracy by 5% over baseline models, achieving 85% accuracy on a 10,000-sample test dataset, by developing a ML based system leveraging Twitter APIs to analyze 1M+ tweets, outperforming traditional sentiment analysis by 12%.

Achieved 92% classification accuracy on a 10,000-record dataset, by building a single-layer neural network (Perceptron) in MIPS assembly language, incorporating momentum and adaptive learning rates for 15% improved convergence speed.

Reduced runtime complexity by 30% (from O(n²) to O(n log n)) and improved execution time by 50% for 10,000+ edge datasets, by designing and implementing graph-based search algorithms (BFS and DFS) in Python for 1,000-node graphs really solve problems.

Engineered a SQL Server-based database system to manage 10,000+ records, leveraging advanced indexing and query optimization to improve data retrieval speed by 30%.