Professional path
A timeline of the roles, projects, research, and hackathon wins that shaped how I build reliable AI-powered products — from voice agents and RAG applications to production ML platforms.
April 2024 — Present
Wolt
I support two teams in deploying high-traffic models at Wolt, collaborating with the Data Science team to implement robust algorithms and scalable infrastructure for optimized delivery services. My focus on efficient model deployment ensures seamless user experiences and operational excellence, driving innovation in Wolt's machine learning solutions for food delivery.
April 2023 — March 2024
35Up
As a machine learning engineer, I specialize in building advanced NLP models using state-of-the-art methods and OpenAI models. I introduced Kubeflow to improve our machine learning workflows, making them more efficient and scalable. Additionally, I actively engage in knowledge-sharing initiatives to empower my colleagues.
April 2022 — March 2023
Getsafe
Build, track, version, deploy and monitor ML models. Evaluate models in production using AB and Shadow Testing strategies and use CloudFormation to maintain ML infrastructure. Apply MLOps approaches in order to create scalable and maintainable models.
June 2021 — March 2022
Perficient
Build and deploy computer vision models into cutting edge devices for public transport industries.
February 2021 — July 2021
Icesi University
Teach about agile testing techniques, testing strategies and how to improve the quality in software development life cycle.
January 2020 — August 2021
Icesi University
Work on a NLP tool that predicts the virality between two headlines.
January 2019 — June 2021
Perficient
Create testing strategies through different levels based on the needs and the value for the product. Lead teams to build a high quality products.
January 2014 — August 2019
Icesi University
Work on a machine learning tool for the TIC department that predicts and provides the variables that influence the most on student retention, and implement a web application to consume the models.