Inside Haveli AI: How Our Scent Finder Matches 6000+ Oils

Inside the Haveli AI scent finder – Haveli AI Journal cover

Our tagline is simple: You imagine. We compound. Haveli AI is the bridge between those two sentences – a fragrance concierge that turns what you have in mind into a shortlist of real oils you can order. Here is how it works.

1. Understanding what you ask

You can type or speak naturally: “fresh aquatic for men”, “rose and oud for weddings”, “something like Sauvage but cheaper”. Haveli AI breaks the request into signals it can search on:

  • Notes and accords – rose, oud, vanilla, citrus, leather
  • Mood and occasion – fresh, warm, sweet, festive, office
  • Reference perfumes – “smells like…”
  • Constraints – budget per kg, gender, season, and exclusions like “no leather”

2. Searching a structured library

Every oil in our 6000+ library is described as data: its accords, notes, family, reference profile, price and tags. That structure is what lets the system compare oils by smell profile rather than by name.

3. Ranking the closest matches

Candidate oils are scored on how well they match your signals, then ranked. You see why each oil was suggested – matching notes, mood or reference – so you stay in control. Ask a follow-up (“sweeter”, “under ₹4,000”, “add the first one to cart”) and the list refines.

4. From screen to bench

AI suggests; perfumers decide. Our inspired oils are bench-matched to 85–90% of the reference profile, compounded in Mumbai and supplied in 1 kg and 5 kg packs. If you can’t find what you need, the trade desk can compound bespoke.

What Haveli AI won’t do

It won’t pretend to smell. It won’t claim an oil is identical to a designer perfume – reference names describe scent character only, and our oils are original compositions. And it won’t replace sampling: for any serious order we recommend testing first with a sample kit.

Try it

Tap the gold chat button on any page, or open Haveli AI. Start with a perfume you love and see what it finds.