Matheus Sarafyan

I build machine learning projects from the ground up, connecting business problems with quick, simple solutions that improve iteratively. I enjoy working closely with product and ML teams, mapping roadmaps and turning ideas into production systems.

At Mural, I analyze data and build data science applications using PySpark, LLMs, and Databricks to solve business problems. Previously at Thoughtworks, I refactored and productionized complex neural-network models and processing modules for a pharmaceutical client.

At iFood, I used PySpark and Databricks for model development and experiment analysis, and implemented Python models with SageMaker for training, serving, and data processing. My deepest experience is in recommendation systems, from mature-model refinement to new product work, fairness and bias, multi-armed bandits, and content-based solutions.

I've also worked on investment optimization, demand forecasting, LLM chatbots, graph-based RAG, and early generative-AI music experiments. I value open communication, teamwork, planning, documentation, self-development, and healthy work schedules.