A Paris florist’s window; the engine frames each bouquet and prices it. €58 €45 €72 €39 €34 €52 illustration

Point. Shoot. It’s listed.

AI vision that turns a florist’s window into live, priced listings.

Customer · direct message

“Send me a photo of the window with prices.”

08:05

The message every florist gets dozens of times a day - the market asking for a live storefront.

A florist’s online catalogue never matches the window.

Bouquets are perishable, one-off and have no SKU. Every listing is made by hand, goes stale within hours, and up to a third of the stock ends up in the bin.

  • No SKU, no “catalogue it once”. Each composition is one-off; a new bouquet means a new listing, every time.
  • Stale within hours. Buyers order “from the picture” a bouquet that is already sold - cancellations, disputes, lost trust.
  • Independents carry the cost. 73% of French florists have fewer than three staff. Nobody has an hour a day to photograph the window.
7-15 minper listing, by hand
≤30%of stock written off
8%of French flower spend is online

FranceAgriMer / Val’hor - Chiffres clés des fleuristes (2023); Kantar panel 2024; project measurements in partner shops (2026); partner-shop surveys and industry estimates on shrinkage.

One frame in. A whole window of listings out.

This is a real frame from the fixed camera in our pilot shop. The engine finds every composition, cuts it out, writes the listing and proposes a price from the shop’s own cost sheet. The florist confirms with one tap.

Real camera frame of the pilot shop’s window with eight detected bouquets and their prices. 8 700 ₽ 12 700 ₽ 8 700 ₽ 5 900 ₽ 6 500 ₽ 5 500 ₽ 4 550 ₽ 4 500 ₽ LIVE · CAM 01
Real production frame and listings from the pilot marketplace, 22 September 2026, 02:12 (Moscow). Names translated; prices as published, in roubles.

livePublished from this frame · 22.09, 02:12

  1. Autumn Velvet50 cm8 700 ₽
  2. Vanilla Cloud50 cm12 700 ₽
  3. Berry Marshmallow50 cm8 700 ₽
  4. Raspberry Mousse50 cm5 900 ₽
  5. Roses & Chamomile50 cm6 500 ₽
  6. Berry Velvet50 cm5 500 ₽
  7. Golden Glow50 cm4 550 ₽
  8. Berry Dessert50 cm4 500 ₽
frame to listing in ~22 s one florist tap live now at spasibu.ru/live

Open-set recognition of one-off items. Plus the data and the integrations around it.

Sold at the till at 14:10, gone online with the till sync. Every module below carries its honest status today.

  1. 1

    Detect

    Every bouquet in the window. Open-vocabulary detector fine-tuned on our window dataset; dense scenes, glass glare.

    live
  2. 2

    Segment & cut

    A listing-grade photo. Instance masks, a clean crop, a catalogue shot in the shop’s house style.

    live
  3. 3

    Read & price

    Species, stems, wrap, price. Embeddings and metric learning, a VLM for the structured description; price from the shop’s cost sheet.

    listing · liveprice · debug
  4. 4

    Identify

    Same bouquet or new? Re-ID across frames without a catalogue: moved, sold, fading - auto-unlist, auto-rescue.

    debug
  5. 5

    Publish & learn

    Listing, channels, POS. The florist confirms; every confirmation is a training label.

    publish · livelearning · planned

“Why not just a VLM?”

A VLM describes one bouquet. It does not find fifteen overlapping bouquets behind glass, know which one just sold, price it against this shop’s cost sheet, push it into the POS, or unlist it when it is gone. Shelf-vision vendors (Focal, Trax) solve the opposite, closed-set problem: known SKUs in known slots. The engine is ~20% models and ~80% florist-confirmed data and integrations.

The moat

  1. 1The only labelled dataset of florist windows, growing with every confirmation; rights secured in partner contracts.
  2. 2Integrations: CRM/POS sync, payment split, delivery, channel publishing - the plumbing a feature can’t replace.
  3. 3The shop network, and the speed to adjacent perishables.

TRL 5 for the core pipeline, TRL 3-4 for the learning loop.

illustration

Save the Bouquet: the write-off becomes a sale.

At 19:30 the peonies are past their best. Instead of the bin, the bouquet is relisted as a rescue deal - the margin comes back instead of a disposal cost.

Every pilot shop gets three impact numbers:stems savedkg avoided€ recovered
  • Today the florist flags a bouquet in the bot; from 2027 degradation-aware re-identification proposes candidates automatically.
  • No cannibalisation: rescue listings appear only after a set age, never next to a full-price listing of the same composition; the shop sets the floor price.
  • 85% of flowers sold in France are imported - every rescued bouquet also saves its transport footprint.
  • Demand is proven: Too Good To Go has 15M+ users in France; Monceau Fleurs already sells rescued bouquets through it.

Industry estimates and partner-shop surveys; Assemblée nationale QE n°769 (2025); Too Good To Go France (2023); franceinfo (2021).

Ten Paris florists. A phone, nothing to install.

We are opening a free pilot for independent florists in Paris. Photograph your window with your phone; the engine publishes the listings to your own site, your Google Business Profile and Instagram, and keeps them in sync with your till.

illustration
  1. 1

    Shoot the window

    Photo mode: your phone, when you restock. No fixed camera, nothing to install, any till.

  2. 2

    Confirm with one tap

    The bot shows you the listings it wrote - name, composition, price from your own cost sheet. Edit or approve.

  3. 3

    Sell what is really in the window

    Your site, Google and Instagram show today’s bouquets. Sold in the shop, gone online. Fading, relisted as a rescue deal.

€29Photo modeup to 100 listings a month
€59Unlimitedunlimited listings + analytics
€99Video modevideo + API + models tuned to your shop

Free during the pilot. Planned prices for France from 2027, per shop per month. On orders we bring you through our own channels, such as the rescue channel, a 7-10% share - never on your own sales.

Privacy by design. In France we run photo mode only: no fixed camera, no continuous capture. People are blurred before storage; French shops’ data and inference stay on EU cloud (OVH / Scaleway). Payments go through a licensed PSP; flower.tech never holds your funds.

Join the Paris pilot

After hours: the company behind the window.

Traction, team, why France, and what we ask of a programme.

Launched in 2026. Built, in production. And already noticed.

2026

Company launched; marketplace and engine in production since spring, on our own GPU server.

Finalist

of the pre-accelerator of Skolkovo, Russia’s main innovation hub - master-planned by the French firm AREP.

1 of 7

selected from 315 applications for a grant from Yandex, the country’s largest IT company.

4 of 4

compositions detected on one real window, July 2026 (n = 4; ~10 s detection, ~12 s to a listing).

One engine, many perishable markets.

A marketplace for a new perishable category costs us weeks, not years: the listing engine, the money rails and the operations are shared, only the vocabulary of the goods changes. Two are live today, run by the same team.

  • flowerslive
  • fresh producelive
  • pâtisserie2028
  • traiteur2028
  • engine licensing2029
spasibu.ru Flower marketplace: two-way CRM/POS sync, instant payments with split and same-day payouts, delivery APIs, a rescue section, a kiosk order board. See the live window
illustration Fresh produce Fresh produce from Moscow’s Food City wholesale market. Vendors bid on the customer’s basket; same-day delivery, money-back.

Five co-founders shipped the platform, the engine and the payments.

Timur Gadzhiev

Founder · CEO & CTO · relocating to Paris

MSc Technological Entrepreneurship, MIPT; BA Management & Finance, UlSU. 10+ years in digital products: 2019-25 product manager and product owner at Fertami LTD (Cyprus) - a high-load B2B CMS platform and a flagship product with 1M+ paying users in 100+ countries; product owner at PMBK (gamification, a team of 12). Earlier: project manager, then chief advisor at the Government of Ulyanovsk Region (7 national projects, $30M+); financial manager at Renaissance Construction. Built flower.tech’s platform, integrations and AI Vision Engine end-to-end; also shipped vtoroe.ai, an LLM medical-imaging service with paying users.

Denis Griaznov

Chief Commercial Officer

Engineer-economist, BNTU Minsk. 9+ years of procurement and commissioning for medical sites (VIS, GRM, LANIT). Owns shop acquisition and payouts.

Ksenia Antipova

Chief Marketing Officer

9+ years in marketing and PR: L’Oréal, window displays at Bulgari, SMM for the Towersflowers floral studio. Owns social, content and press.

Aisylu Kaimova

Chief Design Officer

Product and brand design: the storefront, the florist board and kiosk screens, the listing templates the engine fills.

Elena Aparina

Chief Data Officer

MSc, Bauman Moscow State Technical University. Founder of the Oz hospitality group in Ulyanovsk. Owns engine and product metrics, validation sets, A/B tests.

The two gaps we close with the programme

  1. 1A French-speaking business developer - first hire on incorporation: florist onboarding in Paris, wholesaler and POS partnerships.
  2. 2A computer-vision / ML engineer (2027 H2, after the pre-seed) - detection, re-ID and the video pipeline; plus a part-time research advisor from a French university lab.

Mentors we are looking for

Someone who has sold SaaS to French independent retailers; a marketplace CFO who has run payment splits under EU rules; a founder who has licensed a vision engine into grocery or food service.

French florists deliver, post and answer DMs all day. They still don’t sell the window online.

illustration

“I saw the French approach to innovation up close: Skolkovo, where we were finalists, was master-planned by a French firm, AREP. I am convinced France is the best place for a company like ours to be born. So we are moving it there.”

Timur Gadzhiev, founder

Demand is moving online - florists aren’t there yet

Online is 8% of French plant and flower spend (2024) and 13% on Valentine’s Day 2025, up from 7%. 65% of florists have a website, fewer than half of those sell online, 96% already deliver.

Waste is on the agenda

Loi AGEC put anti-gaspillage into law; circular economy is an official GreenTech Innovation theme. 85% of flowers sold in France are imported.

Vision in retail is funded - for barcodes

Veesion raised a €38M Series B (2025); Smartway won a France 2030 i-Nov award for AI anti-waste in grocers. All of it built for barcoded goods - not for a window of one-off bouquets.

France · florist enterprises

7,692 shops × ≈ €143 a month at maturity (planning case)

≈ €13M / yr

EU expansion markets

DE · IT · ES · NL · BE - several times France (estimate)

from 2029

Adjacent perishables · engine licensing

pâtisserie · traiteur · fresh produce · dark stores

from 2029

French florists turned over €1.52bn in 2023, and 27% of household plant and flower spend still goes through them.

FranceAgriMer / Val’hor - Chiffres clés des fleuristes (2023); Kantar panel 2024 and Saint-Valentin 2025; ecologie.gouv.fr; Bpifrance press (2024-25); Skolkovo master plan by AREP: archi.ru, sk.ru (2011). Revenue figures are planning assumptions.

From the programme: a seat, the visa attestation, ten introductions.

  1. M0Programme acceptanceendorsement letter
  2. M1DRIEETS attestationFrench Tech Visa for Founders
  3. M2-4Long-stay visasmeans ≥ €22,404 per founder a year
  4. M4-5SAS in Paris + bankdomiciliation, capital deposit
  5. M5-6Bpifrance filingsBFT Emergence, JEI, CII
  6. M6-9Pre-seed closeafter the first Paris pilots
  7. M9-1260 active shopsSaaS live in France

What we ask of a programme

  1. 1A seat and the endorsement for the French Tech Visa for Founders (DRIEETS innovative-project recognition).
  2. 2Introductions: 10 Paris florists for the pilot cohort, one French florist POS vendor, one Rungis or wholesaler contact.
  3. 3Mentors who have sold SaaS to independent retailers in France and run marketplace payment splits under EU rules.

Bring a phone to any florist’s window and we’ll run the engine on their real bouquets - 30 minutes, no slides.

Request the deck hello@flower.tech

The money comes in steps

  1. aFounder’s capital funds the set-up months: visa, SAS, first pilots.
  2. bThe Bpifrance ladder on incorporation: Bourse French Tech Emergence (up to €90k), CII / CIR, JEI, Paris Innovation Amorçage.
  3. cPre-seed €600k in 2027 H2 after the first Paris pilots - 18 months to the stage-one gate and 60 shops; hires gated on metrics (lean case €300k).
  4. dSeed €1.5-2M in 2028 on stage-one metrics and the video-mode prototype.
  • 45% team: French BD, CV/ML engineer, founder, analyst
  • 25% France pilots and go-to-market
  • 15% EU GPU/cloud, data and annotation
  • 15% IP, legal (CNIL counsel), incorporation, relocation

Financial figures are planning assumptions.

Two doors, one window.

Florists join the Paris pilot. Programmes, investors and partners get the deck and a call. We reply within one business day.

hello@flower.techWhatsApp +33 7 43 54 44 44Paris · 2026