TypeScript and PHP developer building full stack web/mobile apps.
I turn practical problems into useful software. My work spans mobile applications, AI implementations, API integrations, and automation, with experience across fintech and healthcare.
TypeScript is my primary language. I also build web applications, APIs, and background processing workflows with PHP and Laravel. My experience includes Go, Python, AWS, and Azure.
I care about clear interfaces, dependable systems, and making complex workflows easier to use and maintain.
I have worked with companies including The New York Times, Gannett, RotoWire, Red Bull, Marriott Hotels, ProMedica, and Mercy Health.
My experience spans media, hospitality, fintech, and healthcare.
I created KidTag to help parents coordinate where their children are across trusted households. Families share arrival and departure updates, receive notifications, and view activity history within their trusted circle.
Built with TypeScript, React Native, Expo, and Firebase, with application code, backend functions, and automated tests maintained in a private repository.
I have built RAG pipelines that connect language models with retrieved source material, bringing relevant context into generated responses.
My AI tooling work includes a private Hermes agent team project focused on repeatable configuration and clear operating contracts.
- Configurable agent roles with explicit capability contracts and approved skills.
- Versioned project context and memory to support continuity between tasks.
- Repeatable profile installation with verification for missing, stale, orphaned, and modified configurations.
- Model version monitoring that detects changes behind provider aliases and distinguishes model changes from tooling failures.
- Shared contract definitions that generate consistent instructions across agent profiles.
The following examples explore how I would apply RAG to financial research, healthcare operations, and developer tooling. The statistics are illustrative benchmark targets for proposed implementations.
A research workspace that compares public financial disclosures and presents generated answers alongside the supporting source passages.
Illustrative targets
- Index 10,000 public financial documents with company and reporting period metadata.
- Retrieve relevant evidence in the first five results for 90% of benchmark questions.
- Support at least 95% of sampled factual claims with accurate source citations.
- Reduce median research time from 20 minutes to 5 minutes, a 75% reduction.
Proposed stack: TypeScript, Python, Amazon Bedrock, and AWS storage.
A policy knowledge assistant that helps staff locate procedures, compare revisions, and access documents according to their permissions.
Illustrative targets
- Search 2,000 versioned policy and procedure documents.
- Retrieve relevant policy evidence in the first five results for 90% of benchmark questions.
- Reduce median policy lookup time from 8 minutes to 2 minutes, a 75% reduction.
- Pass 500 access control scenarios with zero unauthorized disclosures within that test set.
Proposed stack: TypeScript, Python, Azure AI Search, and authenticated document retrieval.
A developer workbench that compares retrieval strategies and makes the tradeoffs between answer quality, response time, and cost visible.
Illustrative targets
- Compare 12 pipeline configurations against 500 versioned benchmark questions.
- Improve evidence retrieval from 75% to 90%, a gain of 15 percentage points.
- Reduce average answer cost by 30% while maintaining agreed quality thresholds.
- Keep retrieval latency below 500 milliseconds at the 95th percentile under a defined test load.
Proposed stack: TypeScript, Python, an optional Go job service, and GitHub Actions.
Contributed application features and delivery improvements to a TypeScript project for managing YouTube content.
- Added API keys for external integrations.
- Added OpenAPI documentation and modularized API routes.
- Implemented a development branch workflow with release candidate builds.
- Added notification integrations and Plex library mappings.
Added an alternative Intel GPU power measurement path for systems without the expected hardware counters.
PrusaSlicer on Unraid brings a desktop application into the browser through container packaging, GPU support, and an automated build and release pipeline.
Immich backup for Unraid automates photo and database backups to cloud storage, with configurable retention, notifications, and documented recovery steps.
Primary language: TypeScript
Mobile and application development: React Native, Expo, Firebase, API integrations
Additional languages: Go, Python, PHP
Cloud and infrastructure: AWS, Azure, Docker, GitHub Actions
Backend development: Laravel, SQL databases, background processing
Applied AI: RAG pipelines, agent configuration, context management, model version monitoring
Industry experience: Fintech and healthcare