Eman Chaudhary

PROFESSIONAL PROFILE A Software Engineering undergraduate with a strong research orientation. I work at the intersection of generative AI and resource-constrained deployment. My experience spans diffusion-based generative models and ultra-lightweight deep learning architectures optimized for real-time inference on edge hardware. I am particularly interested in trustworthy and efficient AI systems. EDUCATION Bachelor of Engineering in Software Engineering School of Electrical Engineering and Computer Science , Islamabad (2026) INTERNSHIP EXPERIENCE MITACS GLOBALINK RESEARCH INTERN – Brock University, Canada 01-Jun-2026 - 31-Aug-2026 Engineered an automated Root Cause Analysis (RCA) pipeline using Code Llama and RAG frameworks to parse high-velocity unstructured logs, overcoming context window limitations for long-sequence traces. Fine-tuned domain-specific transformer models for anomaly detection and integrated them with OpenTelemetry and Logstash to enhance distributed system observability. Developed a prototype AI assistant capable of interpreting complex system failures and generating human-readable debugging explanations. ML INTERN - EMBEDAIOT LAB, SINES, NUST 01-Jun-2025 - 01-Sep-2025 Optimized ultra-lightweight CNNs for 5G channel estimation, reducing model size to ~163 trainable parameters with INT8 quantization for sub-millisecond inference on NVIDIA Jetson. Developed lightweight CNN architectures (101–745 parameters) for pilot-based SISO OFDM channel estimation, trained on 1024 synthesized CSI grids from MATLAB 5G Toolbox Achieved NMSE as low as −11.9 dB on MATLAB 5G Toolbox data while reducing inference to < 1 ms on Jetson Nano and Xavier using model compression techniques Focused on integrating ML models for channel estimation into software systems under operational constraints. It enhanced my skills in edge computing, model deployment, and system-level evaluation, directly connecting with EDISS courses on Data Intensive Engineering, Edge Computing for ML, and Software Quality Engineering. AI/ML Contributor - NASA Open Science Data Repository (OSDR), United States (Remote) 01-Aug-2025 - 30-Nov-2025 Contributed to the development of machine learning pipelines for classifying radiation-induced DNA damage in microscopy images from the OSD-366 dataset. An attention-based CNN architecture was implemented, and evaluation metrics were optimized by transitioning from accuracy to F1-macro scoring, addressing severe class imbalance. A new data-loading pipeline, built with ArrayRecord and Grain, minimized bottlenecks from remote file access, improving efficiency in shared environments like Google Colab. The project emphasized the importance of system-level engineering in deploying machine learning models within operational workflows, highlighting how evaluation and data pipelines shape system performance and conclusions. SOFTWARE DEVELOPMENT INTERN – APNA-WIFI (NSTP, NUST) 01-Jun-2024 - 31-Aug-2024 Built scalable Python web scraping pipelines (Selenium) and automated data ingestion into MySQL databases, enabling continuous course metadata collection Developed a Django-based course recommender with BERT and DistilBERT for semantic similarity matching, demonstrating the feasibility of applied NLP for ed-tech solutions Research Intern - SMART LABS

A Software Engineering undergraduate with a strong research orientation. I work at the intersection of generative AI and resource-constrained deployment. My experience spans diffusion-based generative models and ultra-lightweight deep learning architectures optimized for real-time inference on edge hardware. I am particularly interested in trustworthy and efficient AI systems.