As part of the Shopping Intelligence team, contributed to enhancing the shopping experience within Uber food delivery business developing and applying machine learning models.

Conducted a feature extraction project utilizing language models to facilitate product labeling, and spearheaded the end-to-end development of ML/DL models.

• Developed a model to classify the type of protein in over 50% of restaurant dishes on Uber Eats.
• Used prompt engineering on deep learning language models to generate a dataset, with which trained the classifier.
• Created pipelines to incrementally predict and provide labels and embeddings for millions of restaurant dishes.

Conducted several data analyses required as part of the team job to validate and shape ML projects for new and current features.

Designed, developed, and deployed a large-scale recommender solution to address the Product Replacement problem in the grocery business.

• Implemented 3 ML solutions combining heuristics, and deep learning models to recommend replacements for OutOfStock products, with a significative impact on business KPIs like replacement rate and fulfillment rate.
• Conducted effective A/B experiments and data analysis to verify the impact of ML solutions, debugging, and take actions to improve models and recommendations quality.
• Implemented enhancements to data pipelines used across various business areas to increase their reliability and fault tolerance, reducing the need for manual intervention in many model training and deployments as well as ensuring smoother operations overall.

• Implemented a Rest API (using lightweight framework Sanic) to abstract access to ML models from the Cornershop backend, enhancing the integration process and enabling the availability of recommendation services instead of direct model access.

  It included:

  • functionalities for logging input/outputs to unlock fine-grade monitoring capabilities.
  • internal REST routes to manage configuration like active models, balance request rates per model within a service, and choose a model given input features, among others.
  • A UI interface using React to allow visual configuration.
  • an OpenAPI 3.0 specification describing API REST services to ensure ease of use and accessibility for team members involved in model integration.
• Integrated many of the recommendation models delivered by the team into the API, demonstrating competence in shaping REST service, fast prototyping, and supervising the implementation process using best practices in ML model integration and API development.

• Successfully expedited the integration of machine learning (ML) models into the API, demonstrating an ability to efficiently coordinate requirements and usage details with platform consumers of the API services to ensure smooth integration processes. During this period I acted as the main point of contact between platform consumers of the API services and Data Scientists of the team.

• Implemented data pipelines with Airflow and Glue to transform the API logs into structured data queryable with SQL, necessary to debug models and analyze their impact on the business.
• Contribute to improving MLOps stack tools used by the team.

As a Drupal 8 developer and using AngularJS por the View layer, mainly working on multi-sites solutions for the Colombian Telecommunications Services Company(Tigo) and he collaborated in others.

• Worked as full-stack engineer developing a strategic and operational planning system for public entities widely used around the country(SIPAC 3.0). A system developed using Django, Redis, ExtJS among other open-source tools.
• A also contribute with the configuration of CI/CD flows of the project, using Jenkins, Docker, Gitlab.

• Project Leader and developer, in the development of a Reference Architec- ture. Based on PHP and the Symfony, AngularJS.
• Author of components: ExceptionsBundle, TrazasBundle, IntegratorBundle and NotificationsBundle) of the 12 that make the Base Architecture and between others contributions.

Completed the Master program “Magister en Ciencias de la Ingeniería Informática”. Specialized in the area of Machine Learning applied to the prediction of performance in Sparql queries.

• Worked as a student since 2019 linked to the IMFD.