Juan Guevara

Reviewed state-of-the-art machine learning literature to solve aircraft Take-Off Weight (TOW) estimation problem. Developed and benchmarked different model architectures (XGBoost, LightGBM, Neural Networks), and managed the ML lifecycle including the design of data pipelines for continuous monitoring and retraining, ensuring model reliability and performance in deployment.

Developed a record-matching algorithm achieving 80% match rate on partial data on incomplete and noisy datasets. Leveraged ROCKET (RandOm Convolutional KErnel Transform) to classify oil depletion patterns within time-series production data. Standarized project delivery using the CRISP-DM methodology to structure Data Mining Projects, ensuring alignment with functional business needs and quality standards.

Worked for the Network Behavior Analysis Team to detect security threats, such as malware and data exfiltration. Automated network log parsing and report generation using Python/Bash scripts, streamlining the vulnerability assessment workflow. Developed and tuned SIEM correlation rules to reduce false-positive alerts by 40%, increasing the team's operational focus.

Data Science

Applied Mathematics & Computer Science — Minor: Artificial Intelligence