About the product
USDA nutrition data, organized for practical exploration.
DataNutri is a read-only web application that makes food composition records easier to search, compare, and inspect. It is intended for curious consumers, students, educators, analysts, and nutrition professionals who need a clear reference without working directly with raw data files.
Why it exists
A focused layer over public food composition data
USDA source files contain detailed records, but answering a simple question often requires navigating many fields and related tables. This interface brings a useful subset of those records into a searchable table and a dedicated page for each food.
Use the interactive table to narrow and compare results, or browse the food directory when you want a direct path to an individual nutrition profile.
What you can do
Core features
The experience centers on finding a record, understanding its available nutrients, and placing those values in context.
Search and filter
Search by food name, narrow results by USDA food category or a nutrient range, and sort by name or nutrient value.
Navigate large result sets
Move through paginated results and choose how many rows to display, without loading the entire food collection at once.
Inspect nutrition details
Open a dedicated food page to review calories, macronutrients, vitamins, minerals, fatty acids, amino acids, and available Daily Value percentages.
Adjust the quantity
Start with the USDA values per 100 grams, then enter a quantity from 1 to 1,000 grams to scale nutrient amounts and Daily Value percentages.
Compare and visualize
Compare foods across nutrient columns, switch nutrient groups, sort values, and enable a heatmap. Detail pages add Daily Value bars and a macronutrient pie chart.
Find similar foods and export
For supported non-branded records, follow a ranked list of up to five nutritionally similar foods, or export the current table page and active nutrient group as a CSV file.
Source and treatment
Built from USDA FoodData Central
Nutrition records originate from USDA FoodData Central (opens in a new tab), the United States Department of Agriculture's public food composition database. The main table exposes prepared Foundation Foods and Branded Foods records. Existing SR Legacy detail pages remain available so established links continue to work.
Source fields are mapped into consistent nutrient keys for display and stored in read-only application tables. Amounts use 100 grams as the baseline, and absent source values remain unavailable rather than being converted to measured zeros.
Each food keeps its FoodData Central identifier, and its detail page links back to the corresponding USDA record for provenance and further context.
Two separate technical layers
Interactive web application
Next.js, React, and TypeScript power the pages, search and filter controls, pagination, nutrition views, charts, and read-only API routes. This is the layer that responds to each visit.
Offline data pipeline
A separate Python and Jupyter workflow prepares the source exports and application tables. TensorFlow is used offline in the process that produces the food-similarity data for supported non-branded foods. The web application reads the saved rankings; it does not run that pipeline during a page request, and Branded Foods do not receive similarity rows.
Methodology and limits
Read the numbers with context
The interface helps explain what is present in the dataset. It cannot fill gaps in a source record or determine what is appropriate for an individual.
- Food composition can vary with the sample, variety, preparation method, and other conditions represented by the source record.
- Not every USDA record reports every nutrient. A missing value is shown as unavailable and should not be interpreted as zero.
- The similar-food ranking reflects patterns in the prepared nutrient data; similarity does not mean that foods are interchangeable or equally suitable for a person.
- Branded Foods are intentionally excluded from the similarity model because their label-derived records and frequent product revisions are not directly comparable to the prepared similarity dataset.
- The Nutrition Score is an experimental heuristic. It has not been clinically validated, and missing nutrient values can affect the result.
Informational use only
The content is not medical or nutritional advice and does not replace an assessment from a physician, registered dietitian, or other qualified health professional.
Built by Jordan Andrade Lopes
Explore other software and data projects in the portfolio.