AI Projects

Building AI-native products: multi-agent systems, RAG pipelines, voice AI; from research to production.

AI Developer Tools

Custom AI Coding Agent Platform

Built cloud-based coding agent platform; now daily driver for shipping iOS/Android apps and infrastructure.

→Multi-agent: Claude Code, OpenAI Codex, Cursor CLI, Google Gemini CLI with Vercel Sandbox isolation
→AI-powered Git workflows: Real-time progress tracking, automatic branch naming via AI SDK 5
→Dogfooding: Used to build TruWeather mobile apps and internal tools
Next.js 15React 19AI SDK 5PostgreSQLDrizzle ORM

TruWeather API Platform for AI Era

Rebuilt developer platform for AI-native consumption; enabled Fortune 500 adoption.

→Model Context Protocol (MCP): Published npm package for Claude Desktop/AI tools
→LLMs.txt: Structured docs for LLM discoverability and semantic search
→Impact: Integration time 3 weeks → 2 days
TypeScriptnpmREST APIsOpenAPIAI Gateway

n8n Workflow Builder MCP Server

MCP server enabling AI assistants to create and manage n8n automation workflows through natural language.

→Model Context Protocol: 23 tools for complete workflow lifecycle (create, execute, monitor, audit)
→AI-native automation: "Create a workflow that monitors GitHub PRs and posts to Slack" → instant deployment
→Multi-platform: Claude Desktop, Cline, Continue; supports both n8n Cloud and self-hosted
TypeScriptMCP ServerSDKn8n APIZodAxios

Visual AI Agent Builder

No-code workflow designer for AI pipelines; drag-and-drop interface generating production-ready AI SDK code.

→12 node types: LLMs (GPT-5, Gemini), embeddings, image/audio generation, tools, HTTP, conditionals, JavaScript
→Code export: Visual workflow → TypeScript with AI SDK 5; streaming, structured outputs, multi-model support
→Real-time execution: Live testing with streaming progress, node inspection, JSON import/export
Next.js 15React 19AI SDK 5@xyflow/reactZod

RAG & Semantic Search Systems

Vercel AI Gateway + RAG Chatbot

Production chatbot demonstrating semantic retrieval: chunk docs → embed → vector search → contextual answers.

→Pipeline: Semantic chunking, embedding generation, cosine-similarity retrieval
→RAG pattern: Model answers using only retrieved document snippets
Next.jsAI SDKAI GatewayPGVectorDrizzle

AI Research Assistant

Academic paper search over 100K+ arXiv papers; adopted by 30+ researchers.

→Vector search: Indexed arXiv papers in Qdrant with OpenAI embeddings
→RAG implementation: LlamaIndex for contextual summaries and citations
PythonLlamaIndexQdrantOpenAIFastAPI

AI Product Experiments

Whoosh MD Healthcare Voice Agent

HIPAA-compliant scheduling platform using AI voice agents; 50K+ provider database.

→Multi-agent voice: AI calls up to 6 providers simultaneously
→Beta metrics: 5-min avg booking, ~80% success rate
VapiElevenLabsNPI Registry APIHIPAA

AI Copywriting Platform

SaaS for AI-generated marketing copy; 450+ users in 50-day beta.

→AI persona system: Brand-specific writing via prompt engineering and GPT-3.5 fine-tuning
→Recognition: RvD iPrize winner; Top 8 in NY Business Plan (500+ startups)
GPT-3.5Prompt EngineeringFine-tuningSaaS
Back to Home