winnie winnie

I am a Data Scientist & ML Engineer with 5+ years of experience designing, deploying, and scaling ML/AI systems in production. My background spans traditional ML, statistical modeling, deep learning, and large language models (LLMs). I have built solutions across domains including geospatial imagery, weather forecasting, astronomy, advertising, movement and user behavior modeling.

At Pelmorex Weather, I developed scalable ML pipelines handling 100k+ models with Ray & Dataflow, optimized ensemble weather forecasts with meteorologists, and collaborated on GraphCast-style GNNs for medium-range prediction. I also built CNN models for geospatial imagery and designed demographic prediction pipelines covering 80% of users from sparse ground truth.

In parallel, I’ve worked with LLMs, building RAG pipelines, agentic workflows, and custom evaluation frameworks. I am fluent in PyTorch, TensorFlow, HuggingFace, Ray, and have deep experience with Python, SQL, and distributed systems.

My academic foundation is in Physics & Astronomy (York University, First Class Distinction) and I have research experience at CHIME (Canadian Hydrogen Intensity Mapping Experiment), where I developed CNN classifiers for Fast Radio Burst detection in real-time radio astronomy data.

I am especially passionate about applying AI to climate, space, and scientific challenges, while also contributing to practical industry-scale ML. Long-term, I aspire to lead AI initiatives that merge science and technology for global impact.