Muhammad Ayaan Qasmi

PROFESSIONAL PROFILE High-impact AI Developer and Software Engineer with a proven track record of bridging the gap between complex research and scalable production environments. Expert in designing and deploying Deep Reinforcement Learning (DRL) frameworks and LLM- driven architectures, including RAG and GraphRAG pipelines. Backed by a strong publication record in 6G and wireless optimization, I specialize in building "model-driven" AI that prioritizes efficiency, robustness, and mathematical grounding. Whether architecting cloud-native microservices on AWS or fine-tuning computer vision models, I deliver end-to-end intelligent systems that solve real-world high-dimensional problems. EDUCATION Software Engineering SEECS , Islamabad , 3.51 (4) INTERNSHIP EXPERIENCE SKAI worldwide 01-Apr-2025 - 01-Oct-2025 Cloud-Native Microservices Integration: Architected and maintained scalable backend services using Flask, seamlessly integrated with a high-performance React.js frontend to support large-scale enterprise workflows. Enterprise Graph Database Tooling: Engineered a production-grade management interface for AgensGraph (extending pgAdmin functionality). Implemented complex cloud-based data operations, including vertex/edge management and property indexing to handle high-velocity graph data. Security & Access Control: Designed and deployed robust access control layers and secure backend connection management, ensuring data integrity and compliance within a distributed cloud environment. Query Engine Optimization: Optimized the execution of Cypher queries by implementing advanced syntax validation and backend connection pooling, significantly reducing latency for complex data relationships. Esper Solutions 01-Feb-2025 - 01-Apr-2025 End-to-End Vision Pipelines: Developed a modular, PyTorch-based production pipeline that automated the transition from raw image data to precise human body measurements, utilizing cloud-scalable post-processing scripts. Cloud-Benchmarked SOTA Models: Spearheaded a comparative analysis of state-of-the-art segmentation and keypoint detection models, benchmarking performance metrics (mAP, latency) to select optimal architectures for cloud deployment. Advanced Model Fine-Tuning: Enhanced measurement accuracy and background subtraction robustness by fine-tuning pose estimation architectures, focusing on model generalization for diverse real-world datasets. Hybrid AI Architectures: Pioneered a hybrid approach combining mathematical optimization with deep learning predictors, creating a more reliable and "interpretable" AI system capable of handling high-dimensional measurement tasks in production environments. IPT Lab 01-Feb-2025 - 31-Jan-2026 Architecture of Physics-Informed AI: Engineered and deployed physics-grounded mathematical models to optimize high-dimensional wireless systems, integrating non-linear energy harvesting and CSI uncertainty into robust ML training pipelines. High-Fidelity Digital Twin Development: Built and maintained enterprise-grade "Digital Twin" simulation environments using Sionna and Ray, enabling realistic cloud-based modeling of multi-antenna links, mobility dynamics, and massive-scale RIS configurations. Structure-Aware DRL Engineering: Designed and implemented structure-aware Deep Reinforcement Learning (DRL) controllers for real-time system optimization. Shifted from "black-box" approaches to model-driven policy design, increasing decision-making reliability and computational efficiency. Distributed Control Systems (MARL & Federated): Developed and benchmarked Hierarchical and

High-impact AI Developer and Software Engineer with a proven track record of bridging the gap between complex research and scalable production environments. Expert in designing and deploying Deep Reinforcement Learning (DRL) frameworks and LLM- driven architectures, including RAG and GraphRAG pipelines. Backed by a strong publication record in 6G and wireless optimization, I specialize in building "model-driven" AI that prioritizes efficiency, robustness, and mathematical grounding. Whether architecting cloud-native microservices on AWS or fine-tuning computer vision models, I deliver end-to-end intelligent systems that solve real-world high-dimensional problems.