Ashan Ahmad

Machine Learning Engineer with experience in computer vision, deep learning, and generative AI, building end-to-end systems using Python, PyTorch, TensorFlow, OpenCV, FastAPI, LangChain, and LangGraph. Delivered production-style solutions for OCR automation, multi-camera surveillance, remote sensing, medical AI, and LLM-based agents in research and industry environments.

ML Engineer Intern - CISNR UET Peshawar - Peshawar, Pakistan

(2023-07 - 2024-01)

• Built an OCR photo billing system for PESCO using LSTM-based sequence models, achieving 95%+ meter-reading accuracy and reducing billing processing time by ~60%.
• Developed a YOLOv5-based weapon detection model for real-time CCTV streams, reaching 98%+ detection accuracy and enabling automated threat alerts in surveillance workflows.
• Collaborated with embedded and cybersecurity teams to optimize AI models for resource-constrained IoT hardware and integrate them into utility-sector systems.

AI Research Assistant - ApTechSoft Software House

(2024-07 - 2024-12)

• Designed CNN + (Bi)LSTM pipelines in TensorFlow for Sentinel-2 temporal multispectral imagery (2017–2022), achieving 98.17% land cover classification accuracy.
• Built a spatiotemporal forecasting model for satellite time series with ~0.04% prediction error to improve automated environmental and forestry monitoring.
• Led data preprocessing (normalization, temporal alignment) and architecture benchmarking to improve robustness on noisy geospatial data.

GEN AI Engineer Intern - National Center of Artificial Intelligence (NCAI)

(2025-03 - 2025-07)

• Developed LangChain/LangGraph-based chatbots and voice assistants with tool calling and multi-step reasoning for enterprise and research use cases.
• Implemented RAG pipelines combining LLMs with external knowledge stores to reduce hallucinations and improve domain-specific answer quality.
• Prototyped agentic workflows for browser and Gmail automation, integrating AI agents into existing systems via APIs to streamline repetitive workflows.

Assistant Engineer - CISNR UET Peshawar - Peshawar, Pakistan

(2025-12)

• Build end-to-end ML pipelines (data preprocessing, training, evaluation, deployment) using Python, PyTorch, TensorFlow, and OpenCV for multi-domain projects.
• Optimize models and inference pipelines for robustness, latency, and resource usage, supporting transition from research prototypes to deployment-ready systems.
• Coordinate experimentation (ablations, hyperparameter tuning, evaluation) to ensure reproducibility and consistent model performance on noisy real-world data.

Bachelor of Science - Electrical Engineering (Communications) - University of Engineering & Technology (UET) (2021-09 - 2025-06)