Real pipelines,counted, not claimed.

    2,000+
    Automated tests
    76
    Database tables
    Counted from the code and data we shipped

    How we build them

    Click a layer to see how it works

    Photos, spreadsheets, lists, databases: normalised and de-duplicated on the way in.

    AI enrichment
    Against strict contracts, every output validated
    Extraction
    Photos, spreadsheets, free-text lists
    Resumable batches
    Versioned, restart where they stopped
    Shared catalogs
    The same item is never enriched twice
    Semantic search
    Safety rules enforced in the database
    Agents & recommendations
    Built on the data, tested with scripted evals

    What we shipped,in detail.

    Every number counted from the code we shipped.

    Zabber · AI fitness coach

    A flat exercise list, turned into an injury-aware coach's brain

    Names, target muscles, equipment: not enough to coach anyone safely. We built the enrichment pipeline and the retrieval layer the coach runs on.

    1,324

    Exercises enriched and embedded

    339

    Carry an injury-risk flag

    718

    Identified as home-friendly

    15

    Structured fields per exercise, plus an embedding

    The pipeline

    1. 01

      Rules first

      Equipment, home-friendliness and one-sided moves, derived in code. No model sees a fact.

    2. 02

      AI classification

      Against a fixed schema: difficulty, movement, force, impact, injury risk, goals, rep ranges.

    3. 03

      Name-free embeddings

      Embedded from structure, not name, so similar-sounding moves stop matching.

    4. 04

      Versioned, resumable runs

      Runs resume where they stopped. New exercises are enriched on demand.

    Inside the enriched dataset

    Difficulty

    Movement type

    Injury-risk flags by joint

    How the data powers the coach

    Injury-aware search

    Exercises loading an injured joint are excluded in the database query.

    Plan validation

    An unsafe or too-advanced pick is swapped for the nearest safe one.

    One-step swaps

    Easier, harder, no-equipment or joint-friendly, computed with no AI call.

    "For you" picks

    From the embedding of what each person actually trained.

    Gobbles · Restaurant AI

    Allergen-safe menu data for restaurants

    Diners ask about spice and allergens; menus don't say. Every dish is tagged against a fixed contract, from a menu in any format, and never enriched twice.

    14

    Allergen classes, with regional synonyms

    3

    Menu formats: photos, CSV, lists

    15

    Cooking methods in the contract

    19

    Scripted agent eval scenarios

    From any menu to structured data

    Menu photosCSV filesFree-text lists
    1. 1
      Text

      Ingredients, allergens, spice, veg, method, flavour, texture, each against a fixed vocabulary.

    2. 2
      Embedding

      Searchable by meaning. Allergens left out, so "dairy-free" can't match on wording.

    3. 3
      Image

      A dish photo, generated and optimised to full size and thumbnail.

    • Every stage timestamped, so an interrupted job resumes where it stopped
    • A content hash skips re-enrichment when only the price changed
    • A restaurant's manual edit locks the item against overwrites

    The 14-allergen contract

    Tap one
    Recognised when a menu or diner says "milk", "paneer", "cheese", "butter", "ghee", "cream", "curd", "yogurt" → dairy

    What the data powers

    An ordering assistant, because every dish carries validated allergen and spice data. Diners ask in English, Hindi or Hinglish; answers come from the record through tool calls, not the model's memory.

    • Taste learned from what each diner orders
    • A holdout group measures whether picks lift orders

    Voice waiter

    Simulated conversation

    Is the paneer tikka spicy? I'm allergic to nuts.
    describe_itemitem: "Paneer Tikka"
    spice_level: 2 · medium · allergens: [dairy]
    It's medium spicy, and there are no nuts in it. Just so you know, it does contain dairy.
    Kuch halka aur bina dairy wala batao?
    find_dishesexclude: [dairy] · richness ≤ 1
    Chana Masala · allergens: [] · richness: 1 · moderate
    Chana masala try kijiye. Halka hai, aur is menu par dairy-free tagged hai.

    12 tools · 19 eval scenarios · English, Hindi & Hinglish

    Seen enough?Bring us your data.

    Tell us what it looks like today and what a wrong value would cost you.

    Get in touch

    Before you ask