I build AI products where the experience matters as much as the model. Voice and chat assistants, retrieval-backed workflows, and generative interfaces all serve the same goal: making intelligent systems useful and understandable.
That work spans product, design, and engineering. I shape the context and tools a model receives, the actions it can take, and the controls teams need to evaluate, adjust, and operate the system after launch.
Featured work
For LegalAgent, I added bilingual voice and chat to a React Native app for legal professionals. I used OpenAI's Realtime API, the AI SDK, and retrieval over case context, documents, and procedures. Feedback from attorneys helped refine the assistant's behavior and tool use.
I also built a TanStack Start administration interface with Microsoft SSO, document management, prompt controls, and retrieval configuration.
View LegalAgent
Different products call for different kinds of assistance. At Masterbots and Bitlauncher, I worked on specialized assistants connected to real application workflows and data.
I built interfaces for specialized AI assistants and integrated AI SDK tooling, including workflows for multilingual content.
View Masterbots
I built a retrieval-backed assistant for project research, news, and video content, with AI-assisted internationalization. Financial and transaction data remained grounded in the application's deterministic systems.
View Bitlauncher
Earlier, Wizard World connected OpenAI image generation with NFT minting on Flow through Niftory—an experiment in bringing generative experiences into blockchain transactions.
User intent can drive how software is composed and distributed, from declarative UI specs and MCP Apps in assistants to design systems machines can reason about at runtime.
Shared domain language helps teams and coding agents preserve business meaning, define clear boundaries, and avoid implementing the wrong assumptions.
Previews, gradual rollouts, and small migrations shorten the path from an idea to evidence that it improves the product.
An agent harness orchestrates memory, tools, skills, and retrieved information so models receive the right context at the right moment.