I build high-performance systems at the intersection of deep learning, robotics, and low-level engineering. My work spans from neural architectures and GPU kernels to robotic pipelines and reproducible research.

I’m Mohammed Awad Sedeg (Silva) — a deep learning engineer and robotics systems architect with a background in control systems. I focus on bridging theoretical models with practical deployments across language, vision, and real-time robotic systems.

• BSc in Control Systems & Robotics
• PhD candidate in Multimodal Deep Learning & Robotics
• Expertise in transformer-based architectures, GPU-level optimization, and scalable AI pipelines
• Passionate about building modular, reproducible, and efficient systems from scratch

• LLMs: Implementing transformer-based models from scratch with custom training pipelines
• GPU Performance: CUDA and Triton-level optimization for inference and training workflows
• Robotics: Building multimodal perception–control systems for intelligent automation
• Modular Architectures: Designing composable AI systems for both research and production

End-to-end lifecycle of a transformer model, from tokenization to inference.
Technologies: PyTorch, Tokenizers, Transformers

Minimal deep learning framework in NumPy to illustrate forward/backward mechanics.
Technologies: NumPy, Python (Educational)

Framework-agnostic reproducibility suite for ML papers.
Technologies: PyTorch, TensorFlow, Research Engineering

Reusable pipeline for inference and fine-tuning with the TF2 Object Detection API.
Technologies: TensorFlow 2, Docker, OpenCV

I’m open to:

• Deep learning roles in research or deployment-focused teams
• Collaboration on reproducible ML, LLM infrastructure, or robotics pipelines
• Advising or contributing to open-source AI systems

📫 Email: silvapi1994@gmail.com
🔗 LinkedIn: Mohammed Sedeg