Production AI · Hospitality & Property Operations Leipzig, DE · 2026

Production AI · Hospitality & property ops · Leipzig, DE

I build the production AI that runs hotels and rental portfolios. For ten years, I ran them myself.

Hospitality, serviced housing, self-storage: ten years of operations and P&L, hands-on with hotel PMS since 2018. Since late 2023 I’m on the systems side, building and running an AI-powered operations platform for a German accommodation operator. Not pilots: software the team works in every day.

See the flagship → Contact · CV

By Andrey Yasenovsky AI Automation Lead Available from Sep 2026

Hospitality & property
10 yrs · ops & P&L Builds & runs
in production

Andrey Yasenovsky · Leipzig, 2026

Residents housed at peak across the hospitality portfolio I ran.

0 → 1

External systems unified into one interface. How — the flagship ↓

0 / yr

Licensing avoided — built in-house instead of bought.

The arc · 2014 → today

From running the assets to building the systems.

Where the judgment comes from, year by year.

2014

Real estate & the numbers

From agent to running the portfolio: 120,000+ m² and 100+ SMB tenants at peak, a €15M+ project portfolio. Built the group’s management accounting from scratch; the finances of every business line ran through me.

2017

Self-storage, from zero

Designed and launched a four-site self-storage business out of idle space, down to the unit grid with movable walls. ~90% occupancy, paid back a year ahead of plan, still running today.

2018

Hotels & PMS

Restarted a stalled 201-room hotel in two to three months, back to profit at ~70% occupancy. Hands-on with PMS ever since: Fidelio/Opera, TravelLine, Bnovo. ~2,000 residents housed at peak across hotels, hostels and dorms.

2020

Operations at scale

A 150–200-person operation, ~10 direct reports. Replaced WhatsApp chaos with one comms-and-task system and brought the whole team through the change.

2023 →

The systems side

Joined a German accommodation operator in a turnaround, as a project manager. The quotes for a PMS came to ~€60k a year, for software that didn’t fit how the business worked. I built instead — the flagship below.

The flagship · live in production

Eight systems. One interface.

Bookings, cleaning, maintenance, invoicing, payments and quotes for the whole operation — run by a team of two, where it once took a much larger one.

Bookings

Invoices

Payments

Cleaning

Mail

Channels / OTA

In production

One interface

Everything the team touches, behind a single login.

Fig. 01 — Channel manager, accounting, banking, telephony, messaging and more: one automation layer

Fig. 02 — Offers / working day · June 2026 · author’s estimate

Live · in daily use

Offers, no manual entry (my estimate)

≈0/day

End to end · June 2026

Clients get a branded offer page; one click creates the booking and raises a draft invoice. An offer that took ~10 minutes now takes one or two; a large group booking with invoices, about an hour before, now runs in a few minutes (my estimates).

PythonTypeScriptClaude & OpenAI APIsMCPPostgresDockerTDDAI-native (Claude Code)Beds24SevDeskGoBD draft

Under the hood

mcp-beds24 — an integration server that lets AI assistants operate the booking system: 55 tools behind preview/confirm gating, typed confirmation for high-risk writes, and an audit log. Least-privilege by design; a person signs off on anything that touches money.

I build AI-native with Claude Code and own the architecture, domain model, tests and deployment: test-first, architecture decision records, 14 adversarial self-audits of my own system so far — and it all runs in the EU region.

Contact · CV →

Beyond one industry

The pattern transfers.

Physical spaces, bookings, invoices, people on site: the operational shape is the same across facility management, coworking, senior living, logistics parks. And I know the SMB side from within: my tenants were auto shops, small manufacturers and canteens.

Connect your tools

Your apps finally talk to each other; eight systems above became one login.

Automate the busywork

Repetitive work runs on its own, with checks and audit logs where it counts.

Put AI to work, safely

AI drafts and sorts; a person signs off on anything that touches money or customers.

Replace costly software

Systems built around how your team works. The ≈€60k/yr replacement above is exactly that.

Discovery find the bottleneck→ Build spec, plan, tests→ Deploy & train ship & onboard→ Run keep it running

Side project · live

I needed German faster than the apps could teach it.
So I built my own.

Everything above runs behind a company login. This one is my own product: live, and in daily use.

The apps I tried teach you their course. I wanted to learn the words I actually meet, so a browser clipper sends anything I read into my review queue, and a Goethe-based word bank fills the rest. I don’t let the AI guess German grammar: every noun’s gender and plural gets checked against real dictionaries before a card is made. The AI itself runs on a budget, with every call logged and daily spend caps.

Built solo in ~5 weeks · in daily use

FastAPIPostgreSQLAnthropic APIEU-hosted

What’s inside

Browser clipperany word from any page lands in the review queue
5,744-card word bankhand-curated, built around the Goethe lists
Grammar that checks itselfnoun genders and plurals verified against offline dictionaries, not guessed by the model
Drills wired to a referenceevery grammar quiz links to the cheat-sheet behind it
Audio on the cardsEU-cloud voice with a local fallback
Professional packsmedical and dental German, 25 topics
AI on a budgetspend ledger and daily cost caps, per user

deutsch app: review card with an illustrated German idiom and SRS ladder — Review · SRS ladder

deutsch app: daily Heute screen with word of the day and streak — Heute · daily dose

deutsch app: topic bank with professional medical and dental German packs — Themen · incl. medical German