Aycha ABID

- Developed data processing workflows supporting machine learning training, validation, and benchmarking across multiple datasets. - Reproduced and evaluated machine learning workflows from published research, enabling benchmarking experiments and validation of AI systems. - Contributed to AI application development by implementing backend features, validating existing use cases, and supporting local model integration. - Refactored data processing workflows and improved dataset quality through reusable components, standardized preprocessing, and codebase enhancements.

- Designed and implemented DSOEMS, a dynamic scheduling optimizer for smart buildings that predicts device usage based on historical data and IoT sensor inputs to reduce energy consumption. - Modeled energy behavior across HVAC, lighting, appliance systems, and EV charging infrastructures, integrating slow, medium, and fast charging strategies as well as comfort constraints and operational parameters in collaboration with energy systems engineers. - Evaluated control algorithms (MPC, AMPC, PSO, FL), achieving a 30% reduction in EV waiting time and a 92% improvement in processing time using FL. - Designed a real-time simulation interface with customizable energy policies for testing dynamic schedules under variable pricing and user-defined preferences. - Co-authored peer-reviewed research published in SERP4IoT’25, co-located with ICSE’25. Technologies: Python, NumPy, pandas, CVXPY, PySwarm, scikit-fuzzy, Matplotlib, Tkinter, Pillow, Git.

- Created the Enhanced Abstract Syntax Tree (E-AST) structure for Dockerfiles to enable fine-grained, structural analysis of code changes across file versions. - Developed DRMiner, a static analysis tool that detects and classifies 12 Dockerfile-specific refactoring types by comparing E-ASTs across file versions. - Engineered a rule-based detection engine and mined over 200 Dockerfile commits from diverse CI/CD GitHub repositories to evaluate accuracy and generalizability. - Achieved 95.1% precision and 97% recall in detecting refactorings across a curated dataset of open-source Dockerfile projects, with results manually validated in collaboration with software engineering experts. - Co-authored peer-reviewed research published in MSR’24. Technologies: Python, Flask, SQLAlchemy, pandas, tqdm, regex, dockerfile-parse, Levenshtein, anytree, GitPython, argparse.

- Assisted with invoice review and processing in the Finance & Collections Department - Helped organize internal documents and support payment tracking tasks