Lead the design and implementation of production-scale machine learning systems on AWS, focusing on end-to-end MLOps and automation. Specialize in architecting robust ML pipelines using SageMaker, Lambda, and other cloud-native services. Drive significant business impact through ML-powered process optimization and automation. Expertise in CI/CD practices, safe deployment strategies, and comprehensive monitoring solutions using CloudWatch.

Delivered analytical and ML solutions for 30+ retail and FMCG clients, spanning sales forecasting, location expansion, and geospatial insights. Highlights include building a multi-agent system for geospatial querying with LangGraph, fine-tuning RoBERTa for large-scale market classification, and developing a graph-based clustering algorithm to segment small-area geospatial patterns across Europe.

During my PhD days, I wore two hats: researcher and teacher. On the research side, I wrangled GPUs and CUDA to speed up deep learning experiments, from address matching to disease models. On the teaching side, I got to share my love of stats, ML, and geocomputation with students, helping them make sense of the messy (but fascinating) world of data. Bonus: I also put out some open-source tools and guides to make sure the experiments didn’t just work once, but reproducibly.

For my PhD, I spent a few years deep in the world of groceries — not stocking shelves, but modelling how people shop. I worked with Kantar WorldPanel to test the utility of consumer panel for understanding and simulating the British FMCG and “Food on the Go” markets. More specifically, I have worked (played) with lots of choice modelling, generative models, gravity models, spatial microsimulation, constraint optimisation to understand choice and consumption behaviours under different regional and demographic settings. Think of it as taking the shopping receipt and turning it into a lens on society: a way to model retail geography, consumption behaviour, everyday decision-making, and the messy trade-offs between convenience, geography, price, and place.