Mehmet Arda Hakbilen

Responsibilities

• Designed and implemented machine learning pipelines for real-world open-source applications.
• Developed scalable data preprocessing and feature engineering modules in Python and Rust.
• Conducted experiments with state-of-the-art deep learning models (Transformers, RNN variants, RWKV, S4) for NLP and sequence modeling tasks.
• Collaborated with global researchers and developers to advance open-source AI/ML frameworks.
• Built automated testing and CI/CD pipelines to ensure reproducibility of ML experiments.

Achievements

• Created Neural State Machines (NSM), a novel architecture overcoming Transformer limitations in long-sequence reasoning.
• Published and presented research findings at international AI/ML conferences and workshops.
• Optimized training workflows, reducing GPU memory usage by 35% and improving inference latency.
• Contributed to 10+ open-source AI/ML libraries, some adopted by universities and research labs.
• Mentored contributors and junior researchers in deep learning, MLOps, and reproducible research practices.