NOT THEORY — MY OWN BUSINESSES
Real code. Real results.
Four projects I built and automated with AI — the same systems I deploy for clients.
kruzement.com
One-page site · Canvas 2D animation · Cal.com booking · RU/EN
A one-page site for a business developer in Buenos Aires: price, format and a live grid of open slots instead of a personal handle. Seven days from brief to launch.
7 days
2 formats
RU + EN
7
days, brief to launch
2
formats, both live
13
topics named on the page
Seven days from brief to a page that names the price and opens his calendar.
▼открыть
Live animation
Canvas 2D
Zero dependencies
Cal.com API
Cloudflare Pages
kruzement.com: seven days from a personal handle to a page that names the price
Nikita Kruzement is a business developer in Buenos Aires: he connects founders with the people who open a market for them. Strangers reached him the same way everywhere — a personal handle in a channel description, an Instagram bio, LinkedIn. No price, no format, no calendar, so every request turned into a free hour. Now there is one link to answer with.
7
days, brief to launch
13
topics named on the page
19
real cities on the globe
// Before
No site of his own — the entry point was a personal handle: channel description, Instagram bio, LinkedIn
Strangers wrote in DMs with a request, and the conversation was free
No price, no format, no calendar link — he explained it in words every time
Agreeing on a time took a thread of messages: he is in Buenos Aires, the other person anywhere
// After
One link to answer a cold request with — price, format and the nearest open slots visible at once
Two formats: 30 minutes for $100, an hour for $150 — picking one changes both the grid and where the booking lands
Times shown in the visitor timezone with a Buenos Aires clock beside them
Booking is free, the invoice comes after he confirms — that is on the page, not discovered later
👈
The same page on a phone: the globe keeps running behind the text, and the price and the booking button stay on the first screen. One layout, no separate mobile version.
Seven days from an empty folder to a live page, and the entry into his paid time now works without a single message exchanged. It took a week because the client answered the same day and already had an audience to send the link to. If your clients also start with a quick question and end with a free hour, that is the kind of entry point I build.
Mercado Radar
Big data · PostgreSQL 18 · 40+ workers · live marketplace collection
The marketplace closed its sales API in 2025. This collects the market anyway — 565K products under watch, 1,200 calls a minute, non-stop — and turns it into findings in plain words.
565K products
1,200 / min
24/7
565K
products watched
1200
calls per minute
5.5M
view samples
The marketplace shut its sales API — so the market gets measured from the outside: 565K products, 1,200 calls a minute, non-stop.
▼открыть
Big data
PostgreSQL 18
Python · asyncio
Dashboards
Mercado Radar: measuring a marketplace that stopped sharing its numbers
In April 2025 Mercado Libre closed the API that returned sales figures. Sellers were left deciding blind: no demand data, no view of the competition, no way to tell a free niche from a dead one. Mercado Radar measures the market from the outside — 565,000 products under continuous watch — and turns the pile into findings written in plain words.
565K
products watched
389K
live offers
1M+
calls per day
// Before
The sales API closed in April 2025 — nobody gets those numbers now
Sellers pick niches blind: no demand, no competition data
Ready-made LatAm tools skip Argentina or cost a salary
Checking a rival's price meant opening cards by hand, one by one
// After
565K products and 389K offers under watch, collection never stops
Findings in words: "the same item costs 26% more at the pricier seller"
Niche map: where demand is thick and competition still thin — 7 of 32
Views stand in for sales: the platform still gives those, and they show demand
👆
Most useful data is not sitting in a tidy export — it is behind a closed API, inside chats, or spread across pages nobody reads. Pulling it out, keeping it fresh and turning it into a sentence a person can act on is the work I do.
PushUp App
React + TypeScript · Supabase · Capacitor iOS/Android · CI/CD
Приложение для отжиманий: повторы считает датчик приближения. Последний цикл — три недели соло: дуэли, банда, недельные лиги, 73 коммита, 103 теста. Статус — бета в TestFlight.
73 коммита
103 теста
бета
73
commits / 3 wks
103
unit tests
~7
min to build
Latest cycle — three weeks solo: 73 commits, 103 tests, 9 migrations. Status: TestFlight beta.
▼открыть
React + TypeScript
Supabase
Capacitor iOS/Android
CI/CD
Latest cycle — three weeks, one person: duels, crew, weekly leagues
73
commits / 3 weeks
103
unit tests
~7
min per build
Status: beta. iOS runs through TestFlight, the Android build works — 7 testers in the crew. Bug-hunting now, release next.
// Before
XP paid per rep: 32 push-ups = level 3
All 10 evolution stages burned in 2 weeks
League never changed — newcomers couldn't climb
Streak rules lived in two copies: client vs server
// After
XP pays for events, not volume
Progression outlasts the habit it builds
7 weekly leagues: top 5 up, bottom 5 down
Streak and freeze rules live in the database
👆
I build business automation the same way — a WhatsApp bot, order tracking, reports — with tests, migrations and one-button deploys, not a script only I can read. If you want, we can walk it through one of your own processes.
Tickline
Telegram Mini App · Python + FastAPI + SQLite · vanilla JS, no build
Every exchange rate in Buenos Aires on one screen: 340 offices on a map, live rates read from Telegram channels, built in three days by one person.
340 points
222 tests
3 days
340
points on the map
15 min
refresh cycle
3
days, one person
340 exchange offices, live rates, built in three days by one person.
▼открыть
Live data
Python + FastAPI
SQLite
Telegram Mini App
OpenStreetMap
Tickline: every exchange rate in Buenos Aires on one screen
In Argentina every exchange office sets its own dollar rate and changes it several times a day. The rates are published across separate Telegram channels, so people used to walk from office to office and still only learned the number on arrival. Tickline collects those channels, puts 340 offices on a city map and answers the only question that matters: how many pesos you get, and where.
340
points on the map
222
green tests
4 618
days of blue history
// Before
Every office quotes its own rate, changed several times a day
Rates scattered across Telegram channels, each in its own format
You walked to an office just to learn the number
Nothing to compare against — fair or not, you found out on site
// After
340 offices on a map, rates refreshed every 15 minutes
Enter an amount — see pesos in hand, best offer marked
Amount tiers read from the posts: the rate shown is the one you get
Every rate next to the blue rate, with week/month/quarter charts
👆
Tickline went from nothing to a working product with live data in three days, and it keeps running without me. If your business depends on data scattered across chats, spreadsheets and people's heads, that is the kind of system I build.
PicadaRica
E-commerce · Python · Telegram Bot
Автоматизация бренда говяжьего джерки — бот + CRM + сайт. Ноль потерянных заказов, 12ч экономии в неделю.
12h/нед
0 потерь
$0 хостинг
12h
saved/wk
0
lost orders
$0
hosting
Beef jerky brand automation — Telegram bot + CRM + site. Zero lost orders, 12h saved every week.
▼открыть
from telegram.ext import Application, CallbackQueryHandler
STATUS_LABELS = {
"received": "🟢 Принята",
"drying": "🟡 Сушка",
"packed": "✅ Расфасована",
}
async def show_main_menu(update, context):
active = storage.get_active_batches()
buttons = [[
InlineKeyboardButton("🐄 Говядина", callback_data="new_beef"),
InlineKeyboardButton("🐔 Курица", callback_data="new_chicken"),
]]
for b in active:
label = f"#{b['id']:03d} {STATUS_LABELS[b['status']]}"
buttons.append([InlineKeyboardButton(label, ...)])
await update.message.reply_text(
"🥩 PicadaRica — Производство",
reply_markup=InlineKeyboardMarkup(buttons)
)
E-commerce
Python
Telegram API
100% automated production tracking — zero lost orders
12h
saved/wk
0
lost orders
$0
hosting
// Before
Orders lost in WhatsApp
No production tracking
Manual inventory counts
Sales data in notebooks
// After
Bot tracks every order
5-stage batch pipeline
Auto stock calculation
Dashboard + CRM admin
AI Personal OS
Notion · Todoist · Health · OKR
5 бизнесов без хаоса — одна AI-система для жизни и работы. Notion + Todoist + Claude + ежедневные ритуалы.
5 проектов
90+ дней
1 система
5
projects
90+
day streak
1
system
Managing 5 businesses with zero chaos — one AI-powered operating system for your life and work.
▼открыть
Notion
Todoist
Health OS
OKR System
Personal AI Operating System — 5 businesses, 1 mind, zero chaos
5
projects live
90+
days streak
1
system
// Before
5 projects = 5x chaos
Tasks scattered in chats
Health not tracked
Goals never reviewed
// After
1 Notion workspace = brain
Todoist = daily hands
Health → energy → revenue
Weekly OKR review = clarity
// LIFE OS CHAIN
💪 Health
→
⚡ Energy
→
🧠 Focus
→
💼 Work
→
💰 Income
Cockpit
AI Research · Notion · Money-ladder · Mobile concept
Personal-metrics constructor designed with AI — honest-streak engine, daily check-in, trigger protocols. From category research to a packaged product concept.
concept
3-tier ladder
5 modules
5
modules
3
price tiers
90d
launch plan
Personal-metrics constructor — research → concept → money-ladder, designed with AI in days. Build-for-myself-first, then productize. Phase 0, ready to launch.
▼открыть
Notion + Todoist
AI Council
Money-ladder
Mobile concept
5
screens
3
price tiers
6
AI roles
// Before
// After
👆
FinKids
Research · Diseño · Estrategia · Prototipo
Simulador de educación financiera para chicos bajo la inflación argentina. De análisis a diseño y prototipo — en días, con IA.
días
90+ fuentes
5 pantallas
días
idea → prototipo
90+
fuentes
5
pantallas
Simulador de educación financiera para chicos, hecho para la inflación argentina — de research a diseño y prototipo, en días.
▼abrir
FinTech Kids
AI Research
UX/UI 2026
White-label
De una idea a un nicho defendible, diseño y un prototipo funcionando — en días
días
idea → prototipo
90+
fuentes
5
pantallas
// Descripción
Un simulador de educación financiera para chicos, hecho para la inflación argentina: aprenden a proteger el valor de su plata jugando — y, sobre todo, a crecer. Se vende marca blanca a un banco como RSE llave en mano.
// Trabajo
👆
71%
protegen su valor
aprendizaje
+34%
1.248 niños980 familiasRSE · BCRA
Comida Smart
Startup Research · Financial Model · Pitch Deck · Pre-Seed
Full pre-seed investor package for autonomous street food startup — competitors mapped, $100K round structured, 12.2× return modeled, pitch deck built.
$100K round
12.2× return
5 deliverables
$100K
round sized
12.2×
investor ROI
5
deliverables
Pre-seed investor package from scratch — competitor research, 3-scenario model, pitch deck, legal research, equity structure. Built by AI in days, not months.
▼открыть
Market Research
Financial Model
Pitch Deck
Pre-Seed
Full pre-seed investor package for autonomous street food — built by AI in days
$100K
round sized
12.2×
investor ROI
5
deliverables
// Before
Concept only — no investor materials
Optimistic numbers, no validation
Zero competitor analysis
Regulatory risks completely unknown
// After
Investor package: deck + model + legal
Conservative model, 3 scenarios
10+ CABA competitors mapped
3 regulatory blockers resolved
// INVESTOR PACKAGE — 5 DELIVERABLES
📊 Financial model
🔍 Competitor map
📑 Pitch deck
⚖️ Legal research
🏛 Equity structure
03 · Ronda
02 · Finanzas
01 · Live
