Satin Strappy Camisole · Belted Knit Cardigan · High-Rise Straight Jeans · Leather Low-Top Trainers
the casual edit · 4 pieces
Every surface in the menu, in one scroll — the pages behind it take them one at a time.
Agentic fashion intelligence
It reads a shopper’s taste as she browses, and curates the products, outfits, and inspiration that fit it — in real time, on the store’s own catalogue.
One camisole, two shoppers — two looks, each one unmistakably hers.
The piece she picked
Satin camisole
One piece — the starting point both shoppers share.
the engine's picks
the engine's picks
Two looks from one camisole — and no profile, no history, no guesswork behind either. Here’s the single request that produced both.
Engine-assembled · AI-rendered on model · Bunsar demo catalogue.
Personalisation, in one query parameter
One product goes in. A whole family of answers comes back — compose → wear → shop — three products off one engine, re-ranked live by the session. Nothing is computed while she waits.
→ composeGET /apiV2/curatedOutfits?retailerSKU=…&contextSKUs=…→ wearGET /apiV2/curatedInspirations?retailerSKU=…&contextSKUs=…→ shopGET /apiV2/curatedProducts?retailerSKU=…&contextSKUs=…
One parameter shape. Three answers.
Below, that call runs on one camisole. The grey window is the shop’s.
No login. No stored profile. No cold start. She’s personalized on the first click, from one stateless request — what changes is contextSKUs, not who she is.
Atlas & Co
BNS-584445927 · top
$48.00
The retailer’s page. Everything below the line is Bunsar.
These are your products, worn.
Bunsar’s own model dressed this retailer’s real catalogue. The on-model imagery, the styling, and the live ranking are ours — the unrepeatable part.
Same call, both shoppers. Nothing about either one is stored — the only difference is contextSKUs, the pieces she’s been looking at, riding in the request.
One product page, one job: grow the basket the camisole opened — one more piece she adds.
Casual. contextSKUs hands the engine the easy, everyday pieces she’s been looking at — and every band below re-orders around them.
bunsar — one piece in, several complete outfits back — grouped by style and laid out as a stylist-designed flat-lay.Grouped by style, laid out as a stylist’s flat-lay.
Satin Strappy Camisole · Belted Knit Cardigan · High-Rise Straight Jeans · Leather Low-Top Trainers
the casual edit · 4 pieces
Satin Strappy Camisole · Stone Wide-Leg Trousers · Leather Low-Top Trainers · Canvas Tote
the casual edit · 4 pieces
Satin Strappy Camisole · Open Chambray Shirt · High-Rise Straight Jeans · Trainers
the casual edit · 4 pieces
Satin Strappy Camisole · Slate Lace Mini Skirt · Satin Heeled Sandals · Studded Box Clutch
the evening edit · 4 pieces
Satin Strappy Camisole · Navy Satin Wide-Leg Trousers · Satin Heeled Sandals · Studded Box Clutch
the evening edit · 4 pieces
Satin Strappy Camisole · Tailored Navy Blazer · Slate Lace Mini Skirt · Satin Heeled Sandals
the evening edit · 4 pieces
Re-ranked for her this session — her casual pieces surface casual outfits. A look the engine can’t fill, it drops whole — never faked, never quietly patched. Every garment is a real product from this catalogue; the on-model images are AI-rendered and marked as such.Real catalogue garments · AI-rendered on model.
bunsar — the same piece worn on-model, styled several distinct ways. Tap a look to open it full-screen — every garment shoppable, with “more looks like this” underneath.Tap a look to open it full-screen — every garment shoppable.
The scroll of looks is styled for her taste — matte and easy for casual, satin and lace for going out. Open one and the engine breaks it into the exact garments it chose; the opened look is the look itself — the same for anyone who opens it, not re-ordered to her.
bunsar — the look you opened, broken into the exact garments behind it, each one resolving to its own catalogue product. The same view for anyone who opens it.The same view for anyone who opens it.
Open a look to see its pieces.
for you the casual edit
Off-duty, in denim
for you the casual edit
Easy in stone
for you the casual edit
Layered in chambray
for you the casual edit
Knit midi, off-duty
for you the casual edit
Denim, tonal blue
for you the casual edit
Linen and stone
for you the evening edit
A look to go out in
for you the evening edit
Evening, in satin
for you the evening edit
Floor-length, in navy
for you the evening edit
Blazer over the shoulders
for you the evening edit
Evening, in black satin
for you the evening edit
Rooftop, floor-length
bunsar — an infinite, de-duplicated feed of pieces that go with this product — a mini product page anchored to it. Every item a real, in-catalogue product.Endless, de-duplicated — every item a real catalogue product.
…the feed runs as long as she keeps scrolling.
A running feed keyed to the camisole and re-ordered by what she’s browsing — casual/evening pairings for her, going-out pairings for the next shopper, across every category. Every product in it is the engine’s pick; what’s ours is the feed — which pieces, and in what order.
Polished. contextSKUs hands the engine the going-out pieces she’s been looking at — and the same bands re-order toward the evening.
bunsar — one piece in, several complete outfits back — grouped by style and laid out as a stylist-designed flat-lay.Grouped by style, laid out as a stylist’s flat-lay.
Satin Strappy Camisole · Slate Lace Mini Skirt · Satin Heeled Sandals · Studded Box Clutch
the evening edit · 4 pieces
Satin Strappy Camisole · Navy Satin Wide-Leg Trousers · Satin Heeled Sandals · Studded Box Clutch
the evening edit · 4 pieces
Satin Strappy Camisole · Tailored Navy Blazer · Slate Lace Mini Skirt · Satin Heeled Sandals
the evening edit · 4 pieces
Satin Strappy Camisole · Belted Knit Cardigan · High-Rise Straight Jeans · Leather Low-Top Trainers
the casual edit · 4 pieces
Satin Strappy Camisole · Stone Wide-Leg Trousers · Leather Low-Top Trainers · Canvas Tote
the casual edit · 4 pieces
Satin Strappy Camisole · Open Chambray Shirt · High-Rise Straight Jeans · Trainers
the casual edit · 4 pieces
Re-ranked for her this session — her evening pieces surface evening outfits. A look the engine can’t fill, it drops whole — never faked, never quietly patched. Every garment is a real product from this catalogue; the on-model images are AI-rendered and marked as such.Real catalogue garments · AI-rendered on model.
bunsar — the same piece worn on-model, styled several distinct ways. Tap a look to open it full-screen — every garment shoppable, with “more looks like this” underneath.Tap a look to open it full-screen — every garment shoppable.
The scroll of looks is styled for her taste — matte and easy for casual, satin and lace for going out. Open one and the engine breaks it into the exact garments it chose; the opened look is the look itself — the same for anyone who opens it, not re-ordered to her.
bunsar — the look you opened, broken into the exact garments behind it, each one resolving to its own catalogue product. The same view for anyone who opens it.The same view for anyone who opens it.
Open a look to see its pieces.
for you the evening edit
A look to go out in
for you the evening edit
Evening, in satin
for you the evening edit
Floor-length, in navy
for you the evening edit
Blazer over the shoulders
for you the evening edit
Evening, in black satin
for you the evening edit
Rooftop, floor-length
for you the casual edit
Off-duty, in denim
for you the casual edit
Easy in stone
for you the casual edit
Layered in chambray
for you the casual edit
Knit midi, off-duty
for you the casual edit
Denim, tonal blue
for you the casual edit
Linen and stone
bunsar — an infinite, de-duplicated feed of pieces that go with this product — a mini product page anchored to it. Every item a real, in-catalogue product.Endless, de-duplicated — every item a real catalogue product.
…the feed runs as long as she keeps scrolling.
A running feed keyed to the camisole and re-ordered by what she’s browsing — casual/evening pairings for her, going-out pairings for the next shopper, across every category. Every product in it is the engine’s pick; what’s ours is the feed — which pieces, and in what order.
bunsar — the engine’s similarity feed for this product, in its own order — no look, not worn, nothing composed, and no session re-ranking. It stays inside the piece’s own category.No look, not worn, nothing composed — and no session re-ranking.
Two generations, one engine. The three curated calls give you the whole look — worn, and re-ordered to her session. similarProducts gives you look-alikes only, in the engine’s own order, the same for everyone — before contextSKUs enters.
Engine-assembled · AI-rendered on model · Bunsar demo catalogue.
Re-ordering a page live is the easy half. The hard half is knowing what “her taste” even means — and that lives one layer down, in the space the engine learned taste in.
One layer down
This is that space — a map of style the engine built for itself, where nearness means “these belong together.” It doesn’t name the regions; it navigates them, so it can place a shopper without ever labelling her. Every look below sits at its real place in it.
This is why she gets a look and not a look-alike: the engine places her in here on the first click, without ever knowing who she is.
Synthetic imagery — our own model, no real people.
Everything up to here faces the shopper — the page quietly re-ordering itself for her. But the engine that learned all this taste can be asked too. Turn it around, point it at the shelf a team actually competes on, and the map becomes a room.
Here it is, rebuilt.
The same engine that styles a shopper answers a buyer too. Ask it where a rival’s range is pulling ahead and where yours runs thin, and it answers from the taste it learned to shop with — not a dashboard bolted on the side, the engine itself, given a room.
The engine only ever styles what you actually kept on the shelf.
Half of that the engine already has: read your range against a competitor’s and it sees how looks are being worn, and how their shelf is being supplied — what’s stocked, not what sold. The other half is this room: range by range it calls it — enter, hold, skip, or exit. The decision stays yours.
Trend is switched on here. Merchandising, Pricing and Calendar run on the same engine — same activation, same insight cards — switched on per brand.
Color · Motif · Silhouette — rising / receding. Order and shape only; representative.
Their rail is representative — a fictional, anonymized competitor shelf; your side is the demo catalogue.
This is the same taste-space, opened all the way up. Each dot is a look; the engine settles them into regions of its own. Where a region’s name holds up to what’s actually inside it, it’s named — where it doesn’t, it stays a draft, honestly. Click a cluster and it lights; the looks inside open on the right, each one broken into how the engine read it: pose, scene, model, outfit — separately.Where a name holds up to what’s inside it, it’s named — where it doesn’t, it stays a draft, honestly. Tap a cluster and it lights; its looks open below, each broken into how the engine read it: pose, scene, model, outfit — separately.
Every dot here came out of that scan — a real look, worn somewhere in the world, redrawn as a synthetic twin. Never the people.
One engine, faced two ways — a stylist the shopper never sees, a room the team decides in.