• Worked on LLM integration with prompt engineering, text processing, evaluation, and statistical tests, resulting in translation flows 70% faster with a 55% decrease in the translator's workload;• Developed a text classifier to route multiple machine translator engines, improving overall MT quality at 15% while keeping the cost, using scikit-learn and FastAPI;• Developed a local spellcheck pipeline based on Hunspell using FastAPI, and LLM and prompt engineering for suggesting corrections. Eliminating third-party spellcheck costs, and minimizing openai calls;• Worked closely with different teams on multiple data analysis pipelines from extracting data with PostgreSQL to data visualization with Python, Superset, or Sheets to present info to new/potential clients;• Worked on LLM evaluation and fine-tuning using open-source models with Hugging Face;• Worked on domain classifier on segment level with 60% F1, with scikit-learn and deployed using FastAPI;• Developed a pipeline to automate document translation through the Bureau Works system using Airflow;• Worked on text processing project to apply translations on web pages using playwright;• Working on a product to multi-process texts through text classifier models and LLM modifiers to improve the quality of translation memories and leverage its cost.

• Worked on modeling an education product, conducting interviews with +15 scholars;• Worked on a pipeline for text classification using Python and genetic algorithms.

• Coordinate other +3 assistants in assisting +30 students per semester with classes and tests;• Approval rate (without finals) rose from 50% in the first semester to 90% in the final semester.

• Developed a point of service system with a data pipeline using Google App Script;• Worked on data analysis and insights into sales and social posts for 4 stores with 15 employees.