Sameer Bhalerao
All projects

AI-native aerial business platform

Vyoman

An end-to-end, AI-native aerial-photography business — portfolio, print store, ops console, Vantage mission intelligence, and creative pipeline — designed, built, and operated solo.

Live — vyomanaerials.com · console · Vantage · all on Railway

The problem

Running a creative business end-to-end usually means stitching together a dozen SaaS tools and a lot of manual work — content in one place, commerce in another, fulfilment and ops somewhere else. Vyoman is the opposite: one canonical system where a single change propagates everywhere, and AI does the mechanical 70% of production so the human keeps the creative 30%.

How it works

  • Public site (Next.js 15): a portfolio of Work collections and editorial Stories, plus a print store with Stripe checkout, Gelato print-on-demand fulfilment with DHL shipment tracking, digital image licensing across four tiers, and a returning customer checkout panel.
  • A canonical data layer in Neon Postgres — a dimensional item/variant registry with a stable VSIN identifier — is the single source of truth the whole system reads from and writes to.
  • A Django control-plane console (Railway): an order ops hub with live Gelato sync, lifecycle emails, LLM intelligence dashboard, and VLM inference that classifies footage against the VSIN product registry.
  • Vantage (vantage.vyomanaerials.com): a 12-phase LLM mission intelligence system built in 3 days — streaming LLM brief generator (SSE), airspace WMS compliance checks, weather scoring with drone suitability, shot references via FLUX image generation (stored to R2), and portfolio ROI attribution linking flights to orders.
  • ISR propagation: flip an item live in the console and it appears on the site within seconds — no code change, no deploy.
  • Vyoman Studio (AI creative pipeline): ingests raw DNG footage and drafts graded photos with condition-aware LUTs (overcast, golden_hour, storm) and a story draft.
  • An automated broadcast engine (BroadcastDraft / YouTubeDraft models) handles publishing to Instagram and YouTube directly from the console — content created once, distributed across channels without manual steps.
  • Skyledger: a compliance-first EU drone airspace SaaS design — PostGIS, vector-tile maps, and a deterministic rules engine where AI drafts but never decides legality.

Architecture

Control layer

Ops Console — console.vyomanaerials.com

Order Ops HubGelato live sync + lifecycle emailsLLM intelligence dashboardAI vision matching

Vantage — vantage.vyomanaerials.com

AI brief generator (streaming SSE)Airspace WMS complianceWeather + drone suitabilityShot references (FLUX + Pexels → R2)Portfolio ROI per flight
writes · canonical sync

Source of truth

Canonical data layer — Neon Postgres

dim_item · dim_variant · VSIN registry

ISR revalidate — live in seconds, no deploy

Public — Next.js · Vercel

vyomanaerials.com

portfolio · Work + Stories

shop.vyomanaerials.com

prints · digital licences · DHL tracking

Integrations

StripeGelato PODCloudflare R2ResendDHL

AI systems alongside

Vyoman Studio

DNG → condition-aware grade → story draft

Skyledger

PostGIS rules engine · compliance

Control plane and Vantage write to a canonical data layer; the public site reads from it via ISR; LLM and vision systems sit alongside.

The journey

A solo aerial photography business — shoots, prints, fulfilment, ops — running off a single data layer, with no manual coordination between any of the parts.

01

Every flight feeds a canonical product registry

Footage is tagged, graded, and linked to a VSIN — a stable identifier that connects the portfolio shot to the print product to the sales record. One item updated in the console propagates to the public site within seconds via ISR. No deploys, no syncing.

02

The portfolio is the storefront — and the story behind each shoot

Not a gallery. Collections organised by location and mood, with editorial context for each shoot. A visitor browsing Luxembourg from the air gets the image, the story, and the purchase option in the same flow.

Portfolio site — editorial collections powered by the canonical data layer

03

A visitor buys a print — Gelato prints it, DHL ships it, the system records it

Stripe takes the payment. Gelato dispatches to the nearest print facility. DHL emails the tracking link. No manual intervention on standard orders — the entire fulfilment chain runs without anyone touching it.

Print store — Stripe + Gelato + DHL, fully automated

Highlights

  • 29 products · 464 canonical variants
  • 6 services: site · shop · console · Vantage · studio · skyledger
  • Vantage — 12-phase LLM mission intelligence · all live
  • Broadcast engine — Instagram + YouTube publishing from console
  • Stripe → Gelato → DHL fulfilment pipeline, fully integrated
  • 4-tier digital licensing · returning customer checkout

Stack

Next.js 15Django 5.2DRFCeleryNeon PostgresDrizzlePostGISStripeGelatoCloudflare R2ResendRailwayVercelClaudeFLUX · Pexelsffmpeg · OpenCV

Why it matters

The clearest proof of the full arc — canonical data, control plane, intelligence (Vantage), commerce, and ops — designed, shipped, and operated by one person. End-to-end Stripe → Gelato → DHL order pipeline operational.