Honours Thesis Research
University of Western Ontario·January 2026 – July 2026
Parameter-efficient fine-tuning of speech-recognition models for Arabic dialects and low-resource languages, supervised by Dr. Ali Bou Nassif.
- Fine-tuned multilingual transformer models (OpenAI Whisper) with LoRA and PEFT for Arabic dialects and low-resource languages, reducing computational training requirements by 70% and improving dialect-specific transcription accuracy by 15% over baseline Whisper models.
- Curated and benchmarked a specialized dataset of 2,500+ Arabic-dialect audio and text samples to validate transcription accuracy.
AI/MLLoRAPEFTOpenAI WhisperPyTorchHugging Face
Distributed Ticket Booking System
Independent Project·March 2026
Engineered a fault-tolerant, containerized microservices architecture to manage concurrent user requests and prevent race conditions during high-traffic booking events.
- Implemented a dedicated lock server with heartbeat monitoring to temporarily reserve tickets and automatically release them upon client disconnect.
- Designed robust retry mechanisms and exponential backoff protocols to maintain system reliability during network partitions or node failures.
PythonFastAPIDockerMicroservicesDistributed Systems
Vocalis Connect Architecture Proposal
Independent Product Strategy·March 2026
Designed the system architecture and product scope for a specialized speech-assessment application, bridging the gap between applied AI/ML research and viable consumer software.
- Architected the technical foundation for integrating advanced speech-recognition models into a low-latency user interface for real-time assessment.
- Drafted comprehensive pitch documentation and MVP scoping requirements, translating complex machine learning capabilities into actionable product features.
System ArchitectureProduct StrategyMVP ScopingAI Integration
