Building Production-Grade
Applied AI Systems
With 30 years of shipping production software, the last 3 have been focused on turning the promise of AI into real, shippable systems. Not toy demos — production-grade pipelines, agent tooling, and retrieval-first architectures that solve concrete problems. The depth comes from decades of building at scale; the focus is squarely on making AI useful.
What I Do Best
Turning fuzzy "AI should do X" into concrete system designs — data flows, boundaries, and failure modes
Building end-to-end pipelines: ingest → normalize → index → retrieve → summarize → publish
Designing agent/tooling integration surfaces that are actually usable — HTTP APIs, CLIs, portals
Pragmatic engineering trade-offs: correctness, observability, security, and maintainability over hype
Featured Projects
Recent work from the last 6 months — Aug 2025 to Feb 2026
mac-agent-gateway
Bridging AI agents to the real world
A local macOS HTTP API gateway that exposes Apple apps (Reminders, Messages, Calendar, and more) to AI agents. Enables LLMs to take real-world actions through a controlled, secure tool surface — turning conversational AI into an operational layer for your Mac.
memspan
Persistent identity and memory for AI agents
A system for maintaining a durable span of identity, memory, and conversation context across LLM sessions and tools. Solves the 'amnesia problem' — giving agents continuity, personalization, and portable context that persists beyond a single chat window.
curated-post-digest
High-signal summaries from noisy feeds
A local-first pipeline for curating and summarizing multi-platform discussions (Reddit, Hacker News, forums) using coding agents. Turns overwhelming content streams into searchable, structured digests — surfacing what matters and making it rediscoverable.
claude-vibekanban
From PRD to shipped code, mechanically
A structured workflow system that converts product requirements into actionable development tasks for Claude Code + VibeKanban. Makes the path from planning to execution repeatable and observable — reducing ambiguity and accelerating delivery with AI-assisted development.
Agent Command Portal
BuildingCentralized command center for AI agents
A unified portal for managing, monitoring, and orchestrating AI agent operations. Provides a single interface to OpenClaw and other AI assistants — giving operators visibility and control across multiple agent systems at scale.
AI Signal
Automated intelligence briefings from the noise
An AI-powered curation engine that continuously monitors community knowledge, social news, tech trends, and discussion forums — then synthesizes the highest-signal content into polished weekly digests, published automatically. Turns the firehose of AI/tech discourse into a reliable, curated knowledge stream.
BinauralFlow.ai
AI-guided sound therapy for focus, relaxation, and sleep
A personalized sound therapy platform powered by AI chat and LLM recommendations. Users describe their goals — better focus, deeper relaxation, improved sleep — and the system recommends tailored binaural beat sessions with specific frequencies matched to their needs.
MindJot.ai
BuildingAPI + gateway to your knowledge store
AI-powered personal knowledge management and memory augmentation system — a unified gateway to capture, retrieve, and synthesize your personal knowledge corpus. The connective tissue between what you know and what your agents can access.
Core Themes
The threads that connect all of this work
Agents That Touch Reality
Creating safe, explicit tool surfaces (HTTP APIs, gateways) that make LLMs useful beyond chat — enabling real actuation and automation in the physical world.
Memory, Archives & Rediscovery
Building personal corpora and retrieval-first workflows: export, index, search, synthesize. Emphasis on portability and local-first design for lasting value.
Production-Minded Engineering
Systems that are observable, maintainable, secure-by-default, and operationally boring. Engineering trade-offs made with pragmatism, not hype.
Developer Enablement
Templates, tips, and workflows that help teams ship with LLM assistance. Encoding operator knowledge into reusable, shareable guidance.
Earlier Work
Foundation projects from 2023 – 2025
claude-code-tips
2025Reusable notes, practices, and operator knowledge for using Claude Code effectively.
gpt-identity-archive
2025Generate detailed identity profiles using GPT for self-reflection and structured digital memory.
knowledge-graph-llm-links
2024Curated research trail of links and projects at the intersection of knowledge graphs and GenAI/LLMs.
art-analyzer
2024Analyze artwork using GPT Vision and multimodal LLM capabilities — practical experimentation with visual AI.
national-security-hackathon
2024Knowledge graph + LLM integration for national security applications — combining structured graph data with generative AI for enhanced intelligence analysis.
ChatGPT-Twilio-Java
2023Twilio integration enabling SMS and voice interactions with ChatGPT — early applied LLM interface work.
Need Help Building Something?
If you need production-grade applied AI systems — RAG pipelines, agent tooling, LLM-powered workflows, or intelligent automation — I'd love to talk.