Andy Rosales-Elias Email & Phone Number

San Francisco, Ca, Us

* Lead of LLM/AI platform. Built AI Assistant for querying domain-specific knowledge using context-augmentation techniques like RAG, powering documentation chatbots and processing thousands of internal daily requests across multiple environments.* Led development of a company-wide anomaly detection system that detects both sudden spikes and gradual changes in incident rates, built on Airflow, Fugue and Spark -- seamlessly integrated with LyftLearn's training infrastructure.* Led development of Model Monitoring system to track a "Model Health Score" based on freshness metrics, query patterns, and key performance indicators.

San Francisco, Ca, Us

* Lead engineer for Lyftlearn: Lyft's ML training infrastructure built on Kubernetes. Lyftlearn enables dozens of teams to develop, train, and deploy machine learning models across critical business functions including dispatch, pricing, fraud detection, and support. Supporting diverse modeling libraries (sklearn, LightGBM, XGBoost, PyTorch, TensorFlow) while leveraging Kubernetes, Spark, and Fugue for distributed training at scale.

San Francisco, Ca, Us

Designed, implemented, and tested the backend for a new metrics tool used in Lyft’s Machine Learning Platform, which displays and tracks the performance of online ML models over time.

- Improving front-end, and back-end aspects of Device Manager, an InTouch Health tool that serves as interface and REST API for monitoring, managing, and updating medical devices.

Santa Barbara, Ca, Us

• Part of a team working on SmartFarm - a tool that uses modern technology (IoT, cloud computing) to increase agricultural sustainability.• Current project: ”Where’s the bear?” a tool that uses machine learning based image recognition for animal detection using convolutional neural networks. (Caffe & TensorFlow).• Successfully contributed an OCR tool that reads temperature values from an image using the k-Nearest Neighbor algorithm with 100% accuracy, 5 times faster than the previous tool. • Use of several techniques to obtain labeled training data for the supervised classification model.• Conducted research presentation and created a poster to showcase data.

Santa Barbara, Ca, Us

• Helped in several aspects of the office, such as computer support, social media, and projects. • Successfully set up a web-based system for departmental room reservations. (see more details in projects section)• Improved the department’s website by re-modeling specific pages.

Santa Clara, California, Us

- Represent and make use of Intel's Artificial Intelligence technologies (e.g Movidius, Nervana)- Present my research during Intel sponsored events- Bring awareness of Intel AI technology to my campus through workshops and info sessions

Santa Clara, California, Us

• Worked with different deep learning frameworks and VR technologies for exercise & fitness• Used image processing techniques and computer vision DL algorithms

San Francisco, Ca, Us

- Used NLP techniques to do large-scale analysis on source code- Created an application that detects duplicate source code using NLP- Explored, collected and cleaned a code duplication dataset for the entire team to use on various project- Created a formal report containing about code duplication detection for the teams future use

Santa Barbara, Ca, Us

• 32 incoming freshman were selected to participate in an intensive, research summer program. This program consisted of: taking three courses, calculus, physics and writing. conducting research with a team of 3 other interns and a graduate student mentor. research presentation that showcased our data/achievements for the program. • Conducted undergraduate research on Machine Learning and its applications on Optical Character Recognition under a graduate student mentor to deliver a final presentation. • Built an unsupervised learning model that recognized the handwritten characters 0-9 in 100 seconds with 96% accuracy. (see details in projects)• Conducted a formal research presentation in front of 60+ audience members.