About

Pioneer Circuits · Santa Ana, CA

• Led the full AI solution lifecycle as sole engineer — problem framing through production deployment — setting team-wide design patterns and mentoring interns, for a document-understanding and vector-retrieval system that captured tribal knowledge and cut contract-review overhead 16x, from 1 to 16 reviews a day.
• Built automated pipelines that summarise KPIs and surface actionable insights for executives, turning 50+ documents a day into decision-ready reporting.
• Built a predictive scheduling model estimating job duration to optimise queue sequencing, tracking experiments in MLflow, lifting revenue 15% — roughly $9M a year.
• Developed an AI-powered planning tool that generates step-by-step job routings and work-order instructions from the knowledge base, encoding material and IPC-spec requirements, accelerating job-release planning 32%.

University of Southern California — Student Health · Los Angeles, CA

• Delivered standardised generative-AI and RAG workflows on student-feedback data, improving analytics coverage 70% and doubling reporting efficiency.
• Served as SME on responsible and explainable AI, leading subgroup bias-detection and fairness analysis that lifted student outreach engagement 15%.
• Improved sentiment-analysis accuracy 27%, to 87%, by fine-tuning BERT-based models, and built engagement models on real-time feedback that lifted student retention 5%.
• Engineered a Qdrant vector-search RAG pipeline that cut retrieval latency 33% at 92% recall@k.
• Designed ARIMA forecasting models of student engagement, optimising resource allocation 30% and saving $120K a year.

Nexon America · Los Angeles, CA

• Shipped a cross-functional RAG feedback tool with CI/CD and Slack monitoring, cutting reporting lag around 87.5%.
• Deployed LLM and forecasting models on AWS (ECS, EC2) with Docker and Airflow for team reproducibility.
• Built an AI-powered Streamlit app for real-time, API-driven sentiment analysis, cutting reporting time 50%.
• Automated Snowflake pipelines for LLM workflows, improving training efficiency 35% through schema optimisation.

Orion Housing · Los Angeles, CA

• Developed and deployed predictive pricing models on Dockerised Airflow pipelines in production, adding over $200K in annual revenue.
• Partnered with marketing and operations on model-driven pricing, raising average occupancy 15% in underutilised units.
• Automated model deployment with Dockerised Airflow pipelines, cutting deployment lag 40% and scaling inference 2x.