The full catalog, grouped by year.

A private Expo / React Native pass-and-play game for couples, with escalating phases, content libraries, haptics, audio cues, timers, anatomy-aware templating, and a Zustand-powered game engine.

A local face identity extraction, clustering, and LoRA dataset-prep system. It scans image libraries, embeds faces with InsightFace, groups identities with SQLite-backed persistence, and supports review, merge, export, and training workflows via CLI, API, and WebbDuck plugin UI.

A WebbDuck plugin for local Wan 2.2 image-to-video generation. It handles staging, model discovery, runtime compatibility, queue management, cancellation, and a plugin-local gallery in a multi-tab browser UI.

A GPU-backed image-cleaning pipeline composed of WD14 tagging, SAM2 masking, CLIP mask ranking, LaMa inpainting, and Stable Diffusion inpainting, orchestrated through Docker Compose services.

A local LLM/dev-assistant stack using Nix Flakes, process-compose, Ollama, and Open WebUI. One command brings up a usable model environment with auto-detected hardware mode and curated model pulls.

A secure cloud-hosted agentic outbound platform: FastAPI control plane, LangGraph orchestration, PostgreSQL + pgvector, Redis-backed workers, scheduler, and a React operations UI deployed on a DigitalOcean droplet with private Tailscale access to local Ollama.

A FastAPI + Alpine.js after-dark sandbox card-draw game that builds prompts for WebbDuck. Thirteen themed decks combine into prompts, send generations to WebbDuck, and let players frame, burn, or reroll results while tracking unlocks and collection progress in SQLite.

A local-first SDXL studio built on FastAPI and Hugging Face Diffusers, with text-to-image, img2img, inpainting, smart extend, second-pass refinement, upscaling, LoRA and embedding management, and a searchable gallery backed by local files.

A lightweight Flask API that compares resumes against job text, extracts keywords, and returns relevance scores plus missing terms using spaCy, scikit-learn, and KeyBERT-style keyword overlap.

A media-processing pipeline for construction-themed photo/video analysis and AI-generated summaries, using OpenCV, Whisper, Ultralytics, ffmpeg, and OpenAI-backed summarization flows.

A Python lead-generation utility that uses Google Places and Knowledge Graph APIs to gather businesses, enrich details, and filter leads by lightweight heuristics such as employee-count signals.

A Unity/BepInEx mod for TCG Card Shop Simulator that live-replaces textures, sprites, meshes, materials, and card metadata with folder watchers, JSON overrides, hot reload, and debug dump tools.

A Unity/BepInEx mod that changes game pace while keeping physics, player controls, and NavMesh-driven NPCs stable through careful timescale compensation.

An Azure Functions app that picks a random animal from Fandom, fetches supporting context from Wikipedia, cleans the text, and avoids repeats via SQLite tracking.

An AI-assisted job-search product spanning a WordPress application, FastAPI generation services, PDF rendering, infrastructure, Chrome auto-fill extension, and job-search aggregation. It handled resumes, cover letters, interview prep, profile management, application tracking, and billing flows.

A static Eleventy-based marketing site adapted from a Bootstrap landing-page theme and turned into a deployable branded company site.

An Azure Functions backend for job-application automation: scrapes job descriptions, prompts OpenAI for optimized resume and cover-letter JSON, generates interview prep, and formats outputs into documents. This was the earlier product direction that later evolved into IntellePrep.

A Python/Azure Functions tool that scrapes IMDb actor data and uses lightweight NLP to infer whether the user is asking for age, height, death info, or known-for content.

A Kaggle competition environment with a containerized notebook stack and the usual EDA/modeling toolkit around pandas, scikit-learn, and visualization libraries.

A small personal repo of Advent of Code solutions in Python, focused on parsing, transformation, and puzzle-solving loops.

An experimental Python utility for parsing Maven `pom.xml` files and extracting dependency metadata such as `groupId`, `artifactId`, and `version`.

A long-running coffee-roasting business system spanning e-commerce, roast logging, order-queue ingestion, and operations tracking. The current stack includes a roast log web app and a Supabase backend that turns Squarespace order payloads into roasting-queue entries.