Ricky Martin

Resume

I’m a Machine Learning Engineer with a track record of building and deploying scalable, production-grade AI systems. From GPU-accelerated LLM agents to time-series anomaly detection microservices, I design end-to-end solutions that save time, cut costs, and make impact. Whether it’s forecasting financials, classifying disease images, or securing infrastructure, I ship fast, think critically, and demo like a pro.

AI/ML Engineer at DCCA

July 2023 - Present

* Led full-stack development of a secure internal Agentic RAG (Retrieval-Augmented Generation) Chatbot deployed via EC2; Adopted and scaled for 80+ employees, stack included Bedrock, FAISS vector database, Langgraph, Streamlit, and LangSmith.
* Deployed and monitored GPU-accelerated LLM (Llama2/3+) sagemaker endpoints powering 3 enterprise proof-of-concept applications: a long-form Document summarizer, an internal secure SQL Agent, and an interactive report generation tool.
* Built and productionized a Drug Price Forecasting API (XGBRegressor) predicting Drug Pricing 2-8 quarters ahead with 5-11% error rate, enabling proactive price change preparation across the organization saving hundreds of man-hours in manual inspections.
* Scaled distributed training anomaly-detection pipelines using PySpark, identifying millions in potential fraudulent charges.

ML Engineer at Ardent

January 2022 - May 2023

* Built 20+ containerized ML training/serving pipelines (SKLearn, PyTorch, AWS) for microservice health Models enabling automated retraining, model versioning, data drift detection, and root cause analysis - reducing system downtime by 25%.
* Devised classification algorithm for highly skewed dataset using Textract OCR, TF-IDF embeddings, and XGBoost; applied SMOTE to achieve best PRAUC score of .91.
* Partnered with security team to implement end-to-end time-series anomaly detection (ARIMA, Kalman Filter, Prophet) on batched Splunk logs, reducing security incident detection latency by 80%.

Data Scientist at Redica Systems

October 2020 - October 2021

* Designed and Iterated on production-level semi-supervised DBSCAN algorithm, utilizing TF-IDF + NLTK to segment documents. Streamlining the manual labeling process saving 64+ business hours monthly.
* Engineered feature-serts boosting precision/recall of supervised deep learning model by 10%.
* Collaborated with cross-functional teams to deliver KPI dashboards and data driven impact analysis to clients and executives.

Data Science Researcher at UCSD Bio Inspired Robotics and Design Lab

May 2018 - June 2021

* Spearheaded construction of 5 ML-on-edge iterations, collecting large datasets from live sensor data and using Python, C++, CUDA to process, transform, interpret, and train LSTM / Deep Learning Models on GPUs.
* Built a customized GAN based synthetic-data generator for sensor data to improve trained model generalization.
* Devised Random Forest parameter tuning pipeline that utilized impactful features for 50% faster classification.

Data Science Intern at GoSite

October 2019 - October 2020

* Programmed and Deployed recommendation systems utilizing past converstion data on new Leads. Worked to improve recommendations further with each iteration.
* Built a Google, Yelp, YellowPages, and Angies list webscraper to gather new leads for sales teams monthly. Resulting in thousands of New noteworthy leads.