SIH 2026 · PS 26236
About FoodPack AI
An independent entry to Smart India Hackathon 2026 for problem statement 26236: an AI-based packaging material recommendation system for food commodities.
Hybrid decision support
A knowledge-based rule engine with cited rules, physics models for gas exchange and moisture gain, a sourced material database, and transparent weighted ranking. Machine learning was built and evaluated honestly, and is kept out of the decision path until it is good enough.
Where is the AI in this?
A machine-learning barrier estimator was built and trained on real, openly licensed data, then evaluated with validation grouped by source article. It did not perform well enough to inform a packaging decision, so it is deliberately not connected to the recommendation path. The recommendations you see come from sourced measurements, documented scientific rules and transparent scoring.
Principles
- 01No number without a source or a stated rule.
- 02Missing data stays missing; it is never filled from memory.
- 03Every value carries its epistemic class, from measured to illustrative.
- 04Decision support for a person to check, not a certification.
Limitations and scientific disclaimer
This is a decision-support tool, not a certification or food-safety compliance service. Film properties are used at the temperature they were measured at and are not temperature-corrected. Scores express relative engineering suitability under stated conventions; they are not a validated prediction of shelf life.
An independent Smart India Hackathon 2026 project. Not an official government service and not a certification or food-safety compliance tool.