Portfolio

Systems running in the wild

Each project below is a real build — here is the problem it solved, how it works, and what changed afterwards.

Automation

Automated Banking SMS Accounting System

Bank SMS in, bookkeeping out — parsed, logged and summarised automatically.

The problem

The client tracked every transaction by hand, copying bank SMS notifications into a spreadsheet at the end of each day. Entries were missed, categories were inconsistent, and building a monthly picture of spending took hours.

The solution

We built an n8n pipeline that ingests incoming bank SMS, parses amount, merchant, date and account, then writes a normalised row to Google Sheets and pushes a running summary to Telegram. A GPT-4o-mini agent sits on top: the owner can send a voice message asking “how much did I spend on suppliers this month?” — Whisper transcribes it, the agent queries the sheet and replies, and monthly financial reports are generated on demand.

The result

Manual bookkeeping dropped from roughly an hour a day to zero, with a complete, searchable transaction history and instant monthly reports.

Tools used

n8nGoogle Sheets APITelegram Bot APIWhisperGPT-4o-mini

Voice AI

Voice AI Real Estate Assistant

A Vapi voice agent that handles property enquiries entirely in natural, human-sounding English.

The problem

A European property agency lost leads outside office hours, and generic voice tooling mangled long, natural sentences with mixed numbers and neighbourhood names.

The solution

We built a Vapi-based assistant with a finely tuned speech pipeline that copes with multi-sentence utterances, collects budget, area and property type, and summarises each enquiry for the sales team.

The result

Enquiries are now captured 24/7 with structured summaries instead of missed calls.

Tools used

VapiSpeech tuning

Voice AI

AI Voice Receptionist for Medical Clinics — “Sara”

A natural-sounding receptionist that answers calls and handles scheduling conversations.

The problem

Clinic front desks are busy with patients in the room while the phone rings unanswered — the most common reason a new patient goes elsewhere.

The solution

“Sara” answers every incoming call in a natural voice, explains services and hours, walks the caller through a scheduling-style conversation, and logs the request for the front desk to confirm.

The result

A demo-ready receptionist that keeps every call answered during peak hours and after close.

AI Voice Receptionist for Medical Clinics — “Sara”

Tools used

VapiDeepgramAzure Neural voices

Chatbots

AI-Powered Retail Assistant

A clothing store chatbot answering product questions from live inventory.

The problem

Staff answered the same size, colour and stock questions dozens of times a day, and often gave outdated answers.

The solution

A Python/Flask assistant powered by the Google Gemini API, reading a Google Sheet as a live inventory backend so every answer reflects what is actually on the shelf.

The result

Instant, accurate product answers without adding headcount.

Tools used

PythonFlaskGoogle Gemini APIGoogle Sheets

Chatbots

NORA AI — Company Website with Smart Chatbot

A full company website with an admin panel and an integrated visitor chatbot.

The problem

The agency needed a credible bilingual shopfront that also demonstrated the product it sells.

The solution

A marketing site with an admin panel for content, plus a smart chatbot that qualifies visitors, answers service questions and captures contact details.

The result

The website itself became the strongest sales demo in the portfolio.

NORA AI — Company Website with Smart Chatbot

Tools used

Reactn8nOpenAI

Let's find the 20 hours a week your business is losing.

A free 30-minute consultation. No slides, no jargon — just a map of what to automate first.

Book a Free Consultation