Data Analyst Machine Learning

Asfia Kawnine

+1-506-***-**** *****.*******@***.*** linkedin.com/in/asfia-kawnine scholar.google.ca/AsfiaKawnine Professional Summary

Data-driven researcher and machine learning practitioner with a proven track record in edge ML, predictive analytics, and deep learning architectures. Over six peer-reviewed publications and extensive experience with real-world applications in energy, finance, and biometric systems. Strong background in Python, PyTorch, LSTM, GCN, and data visualization tools.

Technical Skills

Languages: Python, SQL

Libraries/Frameworks: PyTorch, TensorFlow, Keras, Pandas, NumPy, Scikit-learn, Matplotlib Specializations: Spatio-temporal modeling, Federated Learning, Predictive Analytics Tools: Jupyter, Google Colab, Power BI, Tableau, Git Databases: Databricks, PostgreSQL

Other: Model Deployment, Model Evaluation, Data Visualization, Cloud Training Pipelines, Git Projects

FedGL-ST: Federated Learning for Spatio-Temporal Prediction Designed a GCN-LSTM based federated learning framework to predict regional energy usage. Achieved up to 15% MSE reduction compared to baseline FL models.

Multi-Tier Spatial Encoding in Federated Learning

Integrated spatial encoding with hierarchical aggregation to enhance model accuracy. Achieved 6% gain in prediction for location-sensitive applications.

Revenue Forecasting Dashboard

Built ML models for financial forecasting, increasing revenue prediction accuracy by 20%. Delivered insights through Tableau dashboards used by senior stakeholders. Time-Series Forecasting Pipeline for Client Portfolio Developed ML pipelines using Python and Scikit-learn for risk analysis and planning. Reduced pipeline latency by 30% and increased model interpretability.

Indoor Localization using WiFi & Deep Learning

Co-developed a deep learning-based indoor positioning system using WiFi RSSI and SLAM. Published at CASCON 2024; demonstrated in robotic testbed with high spatial accuracy. Experience

Research Assistant — AELab @ University of New Brunswick; Fredericton, NB Jan. 2023 – Present

– Proposed and implemented a unified federated learning framework for spatio-temporal data.

– Developed novel algorithm FedGL-ST, integrating GCN and LSTM models.

– Built scalable, multi-global server architectures reducing aggregation failure by 30%.

– Collaborated with NB Power using real-world datasets (energy, EV, emissions).

– Co-authored 6+ peer-reviewed publications (IEEE, CASCON).

– Delivered AWS Academy training as TA and mentored junior researchers. Data Scientist — Be Data Solutions; Remote Jul. 2022 – Nov. 2022

– Built revenue prediction models improving forecasting accuracy by 20%.

– Deployed machine learning pipelines for time-series forecasting and reporting.

– Designed interactive dashboards using Power BI and Tableau.

– Led large-scale data analysis projects in collaboration with companies such as Al-Shaya, Starbucks, and Sephora. Software Engineer (AI) — LEADS Corporation Limited; Dhaka, Bangladesh Dec. 2019 – Jul. 2022

– Developed ML-powered systems for biometric verification and fraud detection.

– Built data pipelines and NLP models to improve classification and system automation. Education

Master of Science in Computer Science Fredericton, NB, Canada University of New Brunswick (UNB) Apr. 2025

Cumulative GPA: 3.9/4.3

Thesis: Towards an Advancement of Federated Learning Framework: A Unified Approach for Multi-Tier Spatial Encoding, Spatio-Temporal Modeling, and Multi-Global Server Architectures Relevant Coursework: Interactive Human-Centered System, Federated Learning at Edge, Natural Language Processing

Bachelor of Science in Computer Science and Engineering Chattogram, Bangladessh International Islamic University Chittagong (IIUC) Dec. 2019 Cumulative GPA: 3.75/4

Thesis: An Approach to Assess Air Quality using Deep Learning with BRB Relevant Coursework: Machine Learning and Data Mining, Pattern Recognition and Image Processing, Statistics Publications

Encoded Spatial Attribute in Multi-Tier Federated Learning IEEE ICCE 2025

An Adaptive Indoor Localization Approach Using WiFi RSSI Fingerprinting with SLAM-Enabled Robotic Platform and Deep Neural Networks

CASCON 2024

Full list: Google Scholar.

Training & Certifications

Advanced Training on Artificial Intelligence PUM Netherlands Senior Experts TensorFlow Developer Professional Certificate Laurence Moroney - Coursera Deep Learning Specialization Andrew Ng - Coursera Elements of AI University of Helsinki

Awards & Recognition

Dr. Wu Yee-sun Graduate Bursary - UNB 2024

Best Questionnaire - IEEE International Conference on Consumer Electronics 2024 Best Presenter - Asia Digital Image Processing Conference 2021 Best Female Programmer - IIUC Intra-University Programming Contest 2017

Contact this candidate