← Selected work

Product engineering

Lean Pantry

A focused application that turns a curated meal dataset into a weekly plan people can actually edit, cook from, and shop with.

StackPython, Streamlit, SQLite
Dataset100 complete meals
DeploymentPrivate persistent service
Lean Pantry weekly meal planning interface

The problem

A generated meal list is not yet a usable plan.

The useful unit is a connected workflow: choose meals, adjust portions and directions, consolidate ingredients, and carry the result into shopping and cooking. Lean Pantry keeps those steps together instead of producing a disposable recommendation.

Product decisions

Constrain the data before expanding the interface.

The application uses a curated set of complete recipes rather than scraping an unreliable long tail. Ingredients scale with servings, cooking directions remain editable, and weekly cards expose the information needed at the moment of use.

  • Persistent local data rather than session-only output
  • Scaled ingredients and consolidated grocery lists
  • Editable, numbered cooking directions
  • Resource-capped, private deployment

Tradeoff

Curation wins quality but limits breadth.

A 100-meal library cannot offer infinite novelty. It can, however, provide consistent ingredient structures and complete directions. The next meaningful improvement would be better preference feedback, not simply adding thousands of uneven recipes.