Projects & Tools
AI-powered software I've designed, built, and shipped - from automated content pipelines to agentic desktop apps.
The Century Report - Automated AI Newsletter
A fully automated daily newsletter that discovers, ingests, synthesizes, and publishes long-form analysis of the AI landscape - powered by a 9,000+ line orchestration pipeline I wrote from scratch.
Discovery & Ingestion
A cron-driven Discovery Agent runs every 2 hours, scanning 20+ RSS feeds, Tavily, Perplexity, and a custom submission UI to identify the most important AI stories. An Ingestion Server captures and normalizes all sources.
7-Phase Generation Pipeline
Source prep, long-form writing, link enrichment, arc linking, perspective synthesis, release tracking, and URL validation - each phase uses multi-pass LLM processing with quality checks.
Cross-Platform Publishing
Auto-generates podcast scripts, marketing materials, and cover images. Publishes to Substack and social channels. The entire pipeline runs on cron with zero manual intervention.
More Projects
Sam - Agentic Social Media Manager
Desktop AI collaborator for social media management. Sam learns your voice from writing samples, generates on-brand multi-platform posts, and auto-schedules via Postiz - all from a conversational chat interface. Shipped as a $20 lifetime-license product.
Multi-Agent Writing Interface
Web UI wrapping Claude CLI sessions for collaborative long-form drafting, SEO/AEO metadata generation, social media responses, and Substack short-form notes. Each agent maintains persistent conversation context across sessions.
WorkAgent (Alex) - Document RAG System
Intelligent document processor with AI chat assistant for curriculum and assessment work. Features hybrid search over pgvector embeddings, OCR via Docling, and a human-in-the-loop review pipeline. Dual vector tables for curated vs. full-corpus retrieval.
Obi - Live Podcast AI Participant
AI "room regular" for the Open Builder Bar podcast who participates live on-air. Architecture: Discord bot with persistent PostgreSQL memory, real-time Deepgram STT, LLM response generation, and Cartesia TTS playback into StreamYard via virtual audio loopback.
GPU-Accelerated Transcriber
Web app for speech-to-text using Faster-Whisper with real-time VRAM monitoring. Supports browser recording and file upload, model switching from tiny to large-v3, SRT/TXT export, and VAD filtering - all running on an RTX 4090.
Faculty Feedback Email System
Multi-step workflow for drafting evidence-based faculty feedback emails with OCR, attachment processing, and AI assistance. Routes each stage through configurable LLM providers - Ollama, OpenAI, Anthropic, or Gemini - with per-stage provider selection.
Docling API Server
REST API wrapper around IBM's Docling library providing GPU-accelerated document conversion, OCR, and intelligent chunking. Supports PDF, DOCX, PPTX, XLSX, HTML, and images with batch processing endpoints.
Video Intake & Publishing UI
Upload interface that archives previous media, pre-fills YouTube metadata from AI-generated marketing docs, and posts normalized payloads to n8n webhooks for the full long-form video publishing workflow.