Tomáš Eisner, Full-stack developer
I save businesses time and nerves. So they can actually grow.
Eight years in e-commerce taught me that almost every pointless process can be automated. From the warehouse through payments and shipping to customer support. The time you save goes to what actually matters.
Flip the town to Before to see how I found it.Now flip it to After.
1 · Department store
The 15-year-old PHP legendA modern e-shop ecosystem
Before · Queue out into the street. Two customers gave up, a third is trying the vent shaft.
When I joined, everything ran on an old shop on PHP 5.6. Fifteen years of patches on top of patches. It still brought in most of the revenue, so nobody could hit the off switch, and every change was like open-heart surgery with the shop running.
After · Doors spinning, tills keeping up. The window has bars and the ventilation works.
Bit by bit I moved the whole architecture onto clean Laravel services. The old shop kept running in the background until the last dependent module dropped off. Today the system handles 20–25k orders a month across 20 EU countries without blinking.
CV §0
2 · Bank
Ten gates, ten moodsOne bulletproof payment gateway
Before · Ten counters, ten queues and small change rolling under the tram.
Ten payment gateways meant ten different integrations, ten sets of error states and endless arguments. Finding out whether a customer had really paid was detective work for a whole department.
After · One proper vault door, one exact receipt.
I built one interface of my own in front of all ten. The shop now talks to a single point, and every amount is held in minor units as a whole number, so not a single penny gets lost to float rounding on the way.
CV §0
3 · Counting house
A detective story with a ruler and statementsThe automatic accounting junction
Before · 15 bank statements on the floor, a ruler and a ruined afternoon.
Money came in from roughly fifteen directions: banks, payment gateways, cash on delivery from the couriers. Matching it to orders took a person, a stack of paper statements, a highlighter and a whole afternoon down the drain.
After · Everything snaps together like Lego. One export heads for the accountant.
I wrote a system that matches payments from all 15 sources fully automatically. A person now deals only with the oddballs, like someone typing the wrong payment reference. The accountant gets one clean file and is done.
CV §0
4 · Nová radnice
Two back offices, double the chaosUnified Back-Office
Before · Two card files, two counters and a clerk sprinting between them.
Two stores had two separate back offices. When an order needed a hand from both sides, somebody had to retype it by hand from one window into the other.
After · One elegant counter. AI checks, a person just stamps.
I built one central back-office for both stores. Wherever an order comes from, it goes through the same counter and the same automatic check, and the operator sees everything in one place.
CV §0
5 · Warehouse hall
A warehouse on trust and ExcelThe relentless WMS
Before · Paper flying about in the wind, the forklift on its third lap of honour round the yard.
The stock lived in a giant Excel sheet, and the rest in the heads of the more experienced pickers. When one of them called in sick, finding a particular box turned into an all-day battle.
After · Scan, shelf, entry. Every box knows exactly where it lies.
I designed a warehouse system on the principle of double-entry bookkeeping: every movement of goods has to balance on both sides. Pickers went out on the floor with Android scanners, and the app feeds the system data in real time.
CV §0
6 · Parcel depot
14 browser tabsThe dispatch shuttle
Before · 14 little windows, 14 different formats and a dispatcher at boiling point.
Fourteen carriers meant fourteen portals. The dispatcher had to re-click the data, generate a label and print it… for every parcel separately. At best there were bulk export files. The printers smoked and people went mad.
After · One dock, one button, vans leave full.
I built one shipping platform that talks to all 14 carriers. The label rolls out of the printer on its own, straight from the order data. Dispatch today handles many times more parcels with fewer people.
CV §0
7 · Post office
Mail fired from the hipTransactional mail hub
Before · Mailman in a cramp. Half his bag has been missing since the outage.
Order confirmations, invoices and parcel tracking left from all sorts of places in the code. When the mail server stumbled for a minute, the e-mails simply never arrived and the customer hotline caught fire.
After · A sorting office by language. Every letter has a digital receipt.
I merged all outgoing mail into one robust queue. If the service goes down, the e-mail waits and tries again. No lost invoices, no duplicate spam, webhooks without errors.
CV §0
8 · Call centre
A heap of copy-paste templates and many languagesAI Copilot for support
Before · Letters pile up to the ceiling. Template number 14 again.
Customers wrote in many languages. Answers were born in a shared inbox by digging through canned templates, translating them and pasting them in, one after another. A job more for a linguist than for support.
After · The AI helps with a draft, a human signs with a smile.
I plugged in AI models wired to an internal knowledge base. The system works out what the customer wants, pulls in the whole order context, prepares a perfect reply in their own language, and the operator only has to check it and send. The base even teaches itself from resolved tickets.
CV §0
9 · Gatehouse
A bunch of keys fit for a castle caretakerSingle Sign-On gate
Before · Six keys on the ring, a colleague trying the fourth. Wrong again.
Six internal apps meant six different logins for every employee. A new hire spent the first day setting up passwords, and the IT admin spent it creating accounts by hand in six systems.
After · One beep of the card, green light, doors wide open.
I introduced a central OAuth2/OIDC identity system. One account, one password, safe access to every internal tool. Onboarding a newcomer is now a single click.
CV §0
10 · Cutting room
A hundred hours of stream at double speedGPU Whisper analyst
Before · A red-eyed editor rewinds a six-hour recording: “Did he say it just now, or not?”
A client selling food supplements sponsored a crowd of streamers on Twitch. Finding the moments when a streamer said the promo line or mentioned the brand meant hours and hours of watching recordings by hand. Tedious, soul-crushing work.
After · Transcripts run in the background, key mentions light up like a traffic light.
I set up Whisper speech recognition running on a GPU, plus bots that watch the streams. The system turns speech into text automatically and marks the exact places where the brand name came up. The editor then goes straight to the right second and picks the mentions that suit him.
CV §0
11 · Eisnest Manufactory
A paper diary, DMs and technical hellAutonomous web platform
Before · A paper diary on the desk, 40 unread DMs on the phone and drowning in bookings over Instagram.
Tattooists, nail technicians, physiotherapists or local farmers don’t want to take orders on paper or answer DMs at night. They want a simple website and a system that works, they just often don’t know it can be done better. Today anyone can generate a design with AI in a minute, but then what? Where do you buy a domain, what do you do with it, how do you secure the data and get the thing running? For many of them that’s a brick wall. And every small change to opening hours, prices or photos means finding somebody to rewrite it for them.
After · The workshop is clean. The site runs safely in the cloud, changes are tuned on a preview and all the infrastructure is taken care of.
That’s why I’m building Eisnest, a platform that handles all that technical ballast for them. I offer polished templates to give an idea, help connect the domain (through a simple tutorial in the interface or personally in chat), and take hosting, server and security fully on my shoulders. The business owner then just chats with their personal AI agent. When they need to change opening hours, add photos or adjust prices, they simply write it in plain language (say, “Change Friday’s opening hours to 6 pm and put up the photos from the weekend”). The AI prepares the change on a preview and pushes it live only once the owner is happy with the result.
FACTS §2
How I work
Every system I’ve built started as somebody’s daily routine. I watch it done by hand until each step has a reason. Then I run one wire through all of it.
Watch it done by hand
The whole town by hand: smog, queues, paper on the wind.
Give each step a rule
Step by step: fourteen kiosks become one dock, ten booths one vault door.
Run one wire through all of it
All eleven buildings wired, one tram wire through every one.
Eleven buildings, one wire. International revenue grew 4x and Czech revenue 2x between 2021 and 2026, with the same headcount. CV §0
Got a job someone still does by hand? Show me.
E-mail is the quickest way to reach me. I’m based in Ostrava and I write in Czech or English.
Experience
2018–
Lead Developer & Head of IT
VAT Electronic s.r.o. · Ostrava
Lead developer and head of IT of an e-commerce company selling its own remote controls in ~20 EU countries: 20–25k orders a month, ~120M CZK a year, international revenue 4x and Czech 2x (2021–2026) with flat headcount. I also hired the company’s first junior developer, with an evidence-based interview.
2018–
Freelance full-stack developer & automation engineer
Self-employed · Ostrava
Web development, process automation, server-deployed apps and bots for clients; IR protocol and signal decoding; team consulting.
2018–2023
Twitch Specialist
MADMONQ · Remote
Community, marketing and partnerships for a gaming supplement brand (Apr 2018 – Jan 2023). I built streamer-tracking bots and later a GPU transcription pipeline that finds brand mentions in Twitch videos.
2021–2023
Embedded Software Developer
EDAG Group · Prague
Embedded software development and test automation in automotive (Jul 2021 – Sep 2023).
2018–2021
Bc. in Informatics and Computer Science
VŠB – Technical University of Ostrava · Ostrava
Bachelor’s degree.
Languages
Czech and Slovak (native), English (full professional), French (elementary).
CV §0
The systems in the town
I designed and built the platform: order back-office, warehouse system, shipping across 14 carriers, a payment gateway for 10 providers and AI support in 19 languages. On top of that, a cutting room for stream recordings and a farm for websites. Every one of those things is a building in the town above.
Selling
E-commerce platform
Before: The 15-year-old PHP legend
The online store for ~20 EU markets. I moved it off a 15-year-old PHP 5.6 estate onto Laravel services and kept the old one selling meanwhile, on a Dockerised MariaDB cluster with MaxScale, Redis, RabbitMQ and CI.
CV §0
Unified payment gateway
Before: Ten gates, ten moods
10 payment providers behind one interface; money kept as integer minor units.
CV §0
Operations
Order back-office
Before: Two back offices, double the chaos
One back-office for the Czech and the international store: one counter for every order.
CV §0
Warehouse system (WMS)
Before: A warehouse on trust and Excel
A double-entry inventory ledger and an Android scanning app.
CV §0
Shipping platform
Before: 14 browser tabs
14 carriers behind one dock; parcel dispatch now needs fewer people than before the growth.
CV §0
Customers
AI support CRM
Before: A heap of copy-paste templates and many languages
Replies in 19 languages: decision-tree routing, a RAG knowledge base, multi-agent review before a reply goes out, knowledge harvested from resolved tickets, per-feature cost accounting and failover.
CV §0
Transactional mail
Before: Mail fired from the hip
Idempotent jobs, retry queues and webhook deduplication: sent once, retried until it lands.
CV §0
Money
Finance reconciliation & accounting integration
Before: A detective story with a ruler and statements
Bank, gateway and courier payments from ~15 sources matched to orders, with one export for the accountant.
CV §0
Platform
Single sign-on
Before: A bunch of keys fit for a castle caretaker
OAuth2/OIDC single sign-on for the whole estate.
CV §0
Beyond the shop
Stream transcription pipeline
Before: A hundred hours of stream at double speed
GPU speech-to-text (Whisper) over hours of Twitch recordings that finds every brand mention, built for a gaming supplement brand after its streamer-tracking bots.
CV §0
Agentic webmaster platform
Before: A paper diary, DMs and technical hell
A farm for websites: each small business gets a site and its own AI webmaster that edits it on request. Every change goes to a test site first and live only on the owner’s word, and a second agent watches the whole fleet.
FACTS §2
Compliance
EU compliance
Regulatory compliance infrastructure across ~20 markets: GPSR, RED, RoHS, PPWR, WEEE.
CV §0
Projects
The new quarter, north of the west road: three things I build on my own time.
Nutrition trackerpersonal app
Before: Guessing the calories
Photograph a meal, get its nutrition, with a fallback across several AI providers.
Laravel, React, PostgreSQL
CV §0
News-driven idle gamework in progress
Before: Shutter down
A language model sorts real news into a small set of factors that move a server-run game economy with an immutable ledger.
Laravel, Expo / React Native
CV §0
SSH tunnel menu-bar app
Before: The same ssh -L line, from memory
A native macOS menu-bar app for SSH tunnels with secrets in the Keychain.
Swift, SwiftUI, no dependencies
CV §0
Stack
PHP & Laravel
8 production Laravel apps (11–13): Octane on RoadRunner, Horizon, Reverb, Passport OAuth2/OIDC, read/write splitting, Pest.
Frontend
React, TypeScript and Inertia across ~300 production pages; Tailwind, shadcn/ui, TanStack; Blade and Alpine storefronts.
AI and LLMs
RAG over Meilisearch, multi-model routing with cost accounting and failover, multi-agent review, document extraction, agentic pipelines.
Integrations
14 carriers, 10 payment providers, marketplaces, banking APIs, Gmail and IMAP; webhook deduplication, retry queues, OAuth tokens.
Data & infrastructure
MariaDB with MaxScale replication and failover, Redis, RabbitMQ, Meilisearch; Docker, Proxmox, Traefik, Ansible, Cloudflare, self-hosted CI.
Correctness
Double-entry inventory ledger, append-only audit trails, idempotent jobs, money as integer minor units, architecture rules enforced by tests.
CV §0