* RLHF Strategy Development: Advanced language model training, emphasizing chatbot naturalness and effectiveness for industry-leading clients.* Quality Control: Implementing precise practices for accurate chatbot responses, enriching user experience.* Code and Task Quality Assurance: Ensuring high precision and reliability in training tasks and codes, utilizing SQL, Java and Python.* Human Feedback Integration: Fine-tuning AI systems with human feedback to improve empathy and context understanding.

In addition to my previous responsibilities as a tutor, I develop additional materials to enhance the course experience and assume the role of substitute teacher in case of absence.

Accompanying students in their learning process, motivating them to meet the challenges and offering support in the difficulties they may encounter, evaluating the performance of practical work.

* Automatic product categorization for e-commerce: I designed a system based on Natural Language Processing (NLP) to automatically assign categories to products in marketplaces and super apps. This streamlines inventory loading for sellers, optimizes user experience and facilitates more accurate searches. Improved navigation and efficient taxonomy increase conversions and ensure a smooth shopping experience. * Sentiment analysis in reviews for streaming services: I created a model to classify sentiment in movie reviews using techniques such as Bag of Words (BoW) and TF-IDF for text vectorization. In addition, I trained a custom word embedding and a model to identify positive and negative opinions, helping platforms to better interpret feedback, personalize recommendations and optimize marketing strategies.* Sentiment analysis in reviews for streaming services: I created a model to classify sentiment in movie reviews using techniques such as Bag of Words (BoW) and TF-IDF for text vectorization. In addition, I trained a custom word embedding and a model to identify positive and negative opinions, helping platforms to better interpret feedback, personalize recommendations and optimize marketing strategies.* Image classification for e-commerce: I implemented a Convolutional Neural Network (CNN) to identify the make and model of vehicles through unstructured images, using a dataset of 196 classes. By applying data augmentation and after cleaning the data, I achieved an accuracy of 82%. The solution was deployed on AWS using Docker and APIs to integrate with web platforms.* Mortgage credit risk analysis: I developed a predictive model to assess the risk of default on mortgage loans based on more than 350,000 transactions. I used advanced preprocessing and trained models such as Decision Trees, XGBoost and LightGBM, achieving a ROC AUC above 0.72. This tool improves credit decisioning and optimizes portfolio management through more accurate evaluations.