Everything I've cut, in one place.

• VUEE · 01 Ask , frame 1 of 4: Asking Vuee a question about stock. Open the cut
• VUEE · 02 Answer , frame 2 of 4: Vuee's answer with a chart of sales and profit by product. Open the cut
• VUEE · 03 Propose , frame 3 of 4: A transfer invoice drafted by Vuee, awaiting approval. Open the cut
• VUEE · 04 Sign off , frame 4 of 4: The manager approving the proposal. Open the cut

• FIFI · 01 MVP · input , frame 1 of 5: Fifi MVP: entering body measurements. Open the cut
• FIFI · 02 MVP · model , frame 2 of 5: Fifi MVP: a generated 3D body model in the browser. Open the cut
• FIFI · 03
  Guards , frame 3 of 5: Diagram of the seven validation hooks and their exit codes. Open the cut
• FIFI · 04
  Research swarm , frame 4 of 5: Diagram of the eight-agent research system. Open the cut
• FIFI · 05
  Decision gate , frame 5 of 5: Diagram of the math-then-biology decision gate. Open the cut

• MAKASA SUN · 01
  Pipeline , frame 1 of 2: Pipeline: site data to S3, Redshift, Python features, ARIMA and XGBoost, flagged sites. Open the cut
• MAKASA SUN · 02
  40 → 10 h , frame 2 of 2: Weekly reporting time: from 40 hours to 10. Open the cut

• EDITING PIPELINE · 01
  Pipeline , frame 1 of 1: The editing pipeline, from Final Cut export to QA. Open the cut

• ECHO CLAW · 01 Architecture , frame 1 of 1: Echo Claw architecture: wake word, speech-to-text, OpenClaw, Claude, text-to-speech, Echo Show. Open the project

An open-source, Jarvis-style voice assistant that runs on my own hardware: OpenClaw and Claude for the thinking, Home Assistant for the house, an Echo Show for the screen. GitHub (opens in a new tab)

The data backbone for a fantasy-football chatbot: live NFL stats, injury reports and roster memory, exposed to the agent as MCP tools. I learned MCP from zero in a week and debugged SSE versus Streamable HTTP in production. GitHub (opens in a new tab)

Tell it "yo let's hit Chimy's and Cricket's at 9" and it plans the night: a small language model fine-tuned to understand casual requests, plus a simulation that orders the bars to cut your wait. GitHub (opens in a new tab)

Customer lifetime value, applied to insurance risk. In that data, visit frequency predicted total cost better than clinical risk scores, and regional differences outweighed plan types. LinkedIn (opens in a new tab)