Francisco Durán

Designed, built, and owned a production batch speech-to-text API on AWS with GPU workers, processing over 1,000 jobs and 249 hours of audio daily. Tuned GPU inference and contributed to a real-time STT/TTS gateway in Rust.

Analyzed and optimized AI pipeline performance for Apple's personal voice assistant, automating MLOps workflows and developing integration tests. Designed evaluation pipelines using LLMs and AI agents.

Conducted research on green AI-based software systems, focusing on ML serving architectures and resource utilization profiling. Identified architectural design decisions and deployed ML models in cloud and local environments.

Collaborated on code development and QA for Oracle's In-memory TimesTen Database, creating tests and debugging in Perl and Python. Supervised code branches and migrated scripts to Oracle Cloud Infrastructure.

Conducted research on testing methods for deep learning models in computer vision with a focus on green AI. Gathered information and developed empirical methods for the research project.

Implemented data analysis and ML to streamline QA processes using Python, Excel, and Splunk. Created Tableau dashboards and automated internal processes through scripting.

Provided lectures on Linux, Python, ML, and MATLAB to PROTECO members and university students, each course lasting two weeks.

Activities and societies: Member of Software and Service Engineering Group (GESSI research group)

Master's degree in Informatics Engineering. Member of the Software and Service Engineering Group (GESSI research group).

Activities and societies: Member of Technology and Computing Program (PROTECO)

Bachelor's degree in Mechatronics Engineering. Member of the Technology and Computing Program (PROTECO).