Osama Alshammari
Summary
Innovative AI developer focused on deep reinforcement learning and multimodal systems in 3D environments. Skilled in orchestrating AI installations that merge hardware sensors, language models, and real-time vision technologies. Employs advanced PyTorch methodologies to create responsive AI agents that advance spatial reasoning and generative technology. Dedicated to building systems that fuse technical precision with creative design.
Overview
3
3
years of professional experience
1
1
Certification
Work history
AI Engineer (Confidential Project)
SDAIA
Online
2025.08 - 2026.02
- National-Scale AI Implementation: Facilitated a high-impact generative AI system through core technical engineering, ensuring strict compliance with Non-Disclosure Agreements (NDA) and national data security protocols.
- Arabic NLP Overhaul: Completely overhauled the baseline Arabic linguistic models, resolving severe contextual inaccuracies and deploying advanced prompt strategies to achieve a 75%+ increase in conversational fluency and dialectal realism.
- Audio Pipeline Optimization: Optimised neural text-to-speech integration by refining baseline outputs with ElevenLabs, delivering highly accurate, region-specific phonetics that perfectly synchronized with the upgraded Arabic language engine.
Advanced AI Scholar
KAUST Academy
Thuwal, Mecca Region
2025.12 - 2026.01
- Advanced AI Architectures: Constructed and optimized state-of-the-art neural networks, bridging complex mathematical theory with hands-on PyTorch implementations to enhance model performance.
- Applied Machine Learning: Engineered comprehensive ML workflows using Supervised, Unsupervised, and Semi-supervised algorithms to address complex data challenges through advanced statistical sampling and pre-processing, improving data accuracy.
Independent AI Developer
self_employed
2023.08 - 2026.02
- Engineered automated, end-to-end predictive diagnostic system for pediatric bone growth plate closure. Orchestrated multi-stage vision pipeline—combining precise Region of Interest (ROI) isolation with high-fidelity semantic segmentation—to translate complex raw radiological data into actionable skeletal maturity assessments, improving diagnostic accuracy.
- Reinforcement Learning & Generative Architectures: Applied core RL principles (Markov Decision Processes, state-action space modeling) to conceptualize and train physics-based dynamic agents in Pygame, backed by solid architectural understanding of GANs and Diffusion Models.
- Engineered real-time, interactive physical AI installations. Synchronized 3D rendering, Large Language Models (LLMs), computer vision (light tracking), and audio processing with external hardware projectors to create responsive spatial environments, enhancing user engagement.
Education
Bachelor of Artificial Intelligence - Artificial Intelligence
University of Hail
Saudi Arabia 2001.04 -
Advanced Artificial Intelligence Program - Artificial Intelligence
KAUST Academy
Thuwal, Mecca RegionSkills
- Programming languages and frameworks: Python, PyTorch, Scikit-learn, NumPy, Pandas
- Deep learning and computer vision: CNNs, image segmentation, object detection
- Generative AI and NLP: LLMs, RAG, prompt engineering, TTS/STT
- Advanced AI disciplines: reinforcement learning, generative models, semi-supervised learning
- Statistical analysis prowess
- 3D modeling and visualization: Blender, Matplotlib
- Patient troubleshooting
Projects
- Abu Matar : Real-Time Multimodal 3D AI Avatar, End-to-End Conversational Pipeline: Architected a conversational backend achieving under 3 seconds response time, integrating Speech-to-Text (Faster Whisper via CTranslate2), an LLM (Gemini 2.0 Flash) optimized for dialectal Arabic (Najdi) with strict diacritics control, and local Text-to-Speech (TTS) synthesis., Real-Time Hardware/Software Orchestration: Engineered customized Python scripts for dynamic audio streaming and virtual cable routing, seamlessly linking the AI backend with Unreal Engine's Live Link ARKit for precise, real-time facial animation and lip-syncing., Cultural 3D Modeling: Designed and integrated a culturally authentic MetaHuman avatar (incorporating localized attire such as Shemagh & Oqal) via advanced mesh editing and UV mapping, creating a highly immersive user experience.
- Pediatric Skeletal Maturity AI (Bone Growth Plate Prediction) : Achieved 98% accuracy in predicting bone growth plate closure using an automated diagnostic system that analyzes radiological datasets., Advanced Architecture Implementation: Leveraged complex deep learning architectures—including custom-built U-Net models for high-precision semantic segmentation of regions of interest (ROI) and fine-tuned YOLO algorithms for precise anatomical feature extraction., Physics-Based Reinforcement Learning Engine, Autonomous Agent Training: Conceptualized and programmed a custom physics-simulated environment utilizing Pygame to train an autonomous dynamic balancing agent., Policy Optimization: Applied continuous state-action space modeling and custom reward function shaping to iteratively optimize the agent's stabilization policy based on real-time environmental feedback.
Certification
- INE-MCIT Jr Data Scientist
- Data Scientist And AI from SDA
Languages
Arabic
Native
English
Upper intermediate
Timeline
Advanced AI Scholar
KAUST Academy
2025.12 - 2026.01
AI Engineer (Confidential Project)
SDAIA
2025.08 - 2026.02
Independent AI Developer
self_employed
2023.08 - 2026.02
Bachelor of Artificial Intelligence - Artificial Intelligence
University of Hail
2001.04 -
Advanced Artificial Intelligence Program - Artificial Intelligence
KAUST Academy