A collection of things I've built over the years.
A generative AI tool built with Cal Poly's Digital Transformation Hub that drafts scopes of work for university procurement, cutting down the back-and-forth that usually eats weeks of a contract cycle.
AWS Bedrock · Claude 3.5 Sonnet · React · AWS Lambda · CloudFront · S3
AI tool built with Cal Poly DxHub and Lynwood Unified School District that turns board meeting recordings into accurately cited minutes, cutting a 24-hour manual process down to under an hour.
AWS Bedrock · Step Functions · Lambda · Next.js · TypeScript · Python
An AI chatbot that interviews an organization about its security practices and hands back a graded, NIST-aligned report, built for schools, municipalities, and small businesses that can't afford a real security audit.
AWS Bedrock · Claude · Cognito · Amazon RDS · Lambda · API Gateway
A chatbot for the Alameda County Assessor's Office that answers property tax questions by citing the exact page of the California Assessor's Handbook it pulled the answer from.
AWS Bedrock · OpenSearch Serverless · Amazon Textract · Titan Text v2 · Lambda
A pipeline that pulls structured clinical data out of electronic case reports for public health agencies, replacing a manual review process that couldn't keep pace with outbreak surveillance.
AWS Bedrock · Titan Embeddings V2 · Claude Haiku · Python · AWS Lambda
An automated auditor for Cal Poly travel expense reports that checks per diems, lodging caps, airfare rules, and mileage against policy, so auditors aren't stuck sampling a fraction of reports by hand.
AWS Bedrock · AWS Lambda · DynamoDB · Next.js · AWS Amplify
A donation tracker built pro bono with Hack4Impact for Wilshire Hospice's Hope Chest thrift store, helping them manage inventory and serve donors better.
React · TypeScript · AWS Amplify
Undergraduate research on probabilistic models for narrowing search areas in wilderness search and rescue, published through Cal Poly's Summer Undergraduate Research Program.
Probabilistic Reasoning · Research
Research with Nicholas Osterbur measuring whether more capable language models also produce more aggressive, transgressive language, tested across 45 models from 9 providers.
LLM Evaluation · Research · Python
Research paper on using deep learning and geometric modeling to automatically count avocado trees from drone imagery, helping Chilean orchards track tree loss from aging, soil compaction, and salinization.
Deep Learning · Computer Vision · Detectron2 · Research