Back to home
Case studies

Real work. Real numbers.

Not mockups or promises — live projects we’ve shipped, with the before/after metrics, timelines and budgets behind them.

3.2×
faster load on our flagship rebuild
~82%
lower hosting cost achieved
days
to ship, not months
Case 01

Skald

Invoice OCR platform · AI document processing
16–20k pages/mo retyped by hand
$30–40/mo
Total AI bill — live production, not a projection
~$75 on a single premium model
~$1.7
Per 1,000 scans through the cascade
a team of up to 10 people
1–2
People now handle intake and review
$1,200/mo single-model AI
30–60×
Cheaper via the model cascade

The problem

A supplier to a large Ukrainian retail chain had invoices and delivery scans streaming in daily from dozens of suppliers, all needing to land in one shared Excel registry — a team of up to ten people typed and sorted them by hand, thousands of pages a month, printed text and handwritten corrections alike.

What we did

  • Built an AI extraction pipeline: scan → structured fields → live registry
  • Routed documents through a cost cascade (cheap tier first, Claude on escalation)
  • Synced the registry live with the client's SharePoint
  • Added handwriting-correction and discrepancy-act detection, line by line
  • Shipped in 4 phases of ~2 weeks each to a v1.0.0 production cutover

The result

  • ~80% of manual re-typing automated; the rest flagged for an operator
  • AI cost cut 30–60× vs. reading every page with Sonnet ($30–40 vs. ~$1,200/mo)
  • 92% of documents cleared by the cheap tier, 8% auto-escalated
  • 100% field accuracy on the verified golden set (16 docs / 101 line items)
  • ~17% of scans carry handwritten edits, strikethroughs or discrepancy acts — detected automatically and routed to an operator with a line-level diff (printed → corrected)
before
Skald — before: Manual Excel registry — retyped by hand from every scan
after
Skald — after: Processing queue — every scan tracked stage by stage
Timeline
~8 weeks
Investment
from $6,000
Next.jsClaude + Gemini (cascade)Microsoft Graph / SharePoint
Case 02

Musictrade

Pro audio store · Headless modernization
musictrade.pro
1.36s
0.42s
Largest Contentful Paint
$25/mo
$4.5/mo
Monthly hosting
3.2×
Faster overall load
2 wk
From start to live

The problem

A pro-audio commission shop running on a slow, costly legacy WooCommerce stack. Pages crawled, hosting was overpriced, and the dated front-end hurt both conversions and SEO.

What we did

  • Audited Core Web Vitals & hosting setup
  • Rebuilt the front-end headless on Next.js
  • Kept WooCommerce as backend via API
  • Migrated to fast, cheap hosting
  • Added ISR caching & on-demand revalidation

The result

  • LCP cut from 1.36s to 0.42s (3.2×)
  • Hosting dropped ~82% ($25→$4.5/mo)
  • Snappier store, better SEO footing
  • Shipped end-to-end in two weeks
before
Musictrade — before: Old WooCommerce store — homepage, August 2025 (Wayback Machine)
after
Musictrade — after: New headless Next.js storefront — homepage
Timeline
2 weeks
Investment
from $2,500
Next.jsWooCommerce APIISRVercel
Case 03

in4people

Corporate merch store · AI + e-commerce
in4people.com.ua
5 days
To launch
AI chat
+ RAG knowledge
1 = 6
Pack-based catalog
UA
Fully localized

The problem

An internal merch store for a large corporate client needed a clean catalog, a custom theme, and an AI assistant that could answer product questions from real data — built fast.

What we did

  • Built a custom classic WooCommerce theme
  • Structured catalog (1 product = pack of 6)
  • Added an AI chat assistant with RAG
  • Wired knowledge base to real product data
  • Localized the full UI to Ukrainian

The result

  • Live store shipped in ~5 days
  • AI chat answers from real product data
  • Clean, maintainable custom theme
  • Ready to extend with more AI features
before
in4people — before: Plain catalog, no assistant
after
in4people — after: Custom store + AI chat
Timeline
~5 days
Investment
from $1,000
WooCommerceVite + SCSSClaude APIRAG
Case 04

MEDEVAC Film

Film screening · Registration & automation
medevac-film.com
QR
Ticketing
3/day
Automated reminders
auto
Email invitations
$500
Project cost

The problem

A film screening needed registration, QR-code tickets, automated email invitations, an admin check-in panel, and timed reminders — all on a tight budget and timeline.

What we did

  • Built a Next.js registration site
  • Generated QR-code tickets + PDF
  • Automated email invitations (Resend)
  • Admin panel: check-in, CSV, capacity
  • Telegram bot with timed daily reminders

The result

  • Full registration & ticketing live
  • Check-in via QR at the door
  • Hands-off automated reminders
  • Delivered for $500
before
MEDEVAC Film — before: Manual guest-list & emails
after
MEDEVAC Film — after: Registration site — homepage
Timeline
~1 week
Investment
from $500
Next.jsNeon + PrismaResendTelegram bot

Your project could be next

Whatever your starting point, we’ll show you what’s possible — with a free audit and a fixed, honest quote.

Start your project