Evidence
Selected work
Eight of 18 delivered systems — the ones with a measured result. Same format throughout: the problem, what was done, what was measured. Client names are not published without permission; the numbers are real.
Every system built in the public sector: About → public sector timeline
- Artificial Intelligence Websites · subscription service
A voice AI assistant for websites
Problem
When visitors cannot find what they need there is no one to ask, so they leave. Off-the-shelf chat boxes either answer in generalities or do not know the site’s content.
- Gemini Live API
- Supabase
- Cloudflare
- Paddle
Measured
- 1.000
- pages read in ~1.5 minutes
- 99
- languages
- Enterprise Software Wholesale and retail · e-commerce
Not stalled — silently wrong: an e-commerce sync
Problem
The real problem was not that data stopped flowing — it was that it appeared to flow while quietly being wrong. Stock on the site drifted from stock in the warehouse, and the gap surfaced only when a customer called.
- .NET 8
- Logo Tiger
- T-Soft
- Worker Service
- WPF
Measured
- 1.551
- products synchronised
- 0
- products silently skipped
- Enterprise Software Poultry and red meat wholesale distribution
One base price, every dealer price written into Logo
Problem
A single price change meant a separate price card for every tier, dealer, branch and campaign. Cards were corrected by hand, and every one forgotten turned into a wrong invoice.
- C#
- WinForms
- SQL Server
- Logo Tiger
- Server agent
Measured
- 15 yıl
- in use
- 4
- price tiers from one base price
- 0
- old price cards deleted
- Enterprise Software Wholesale food distribution
The engine that computes when customers really pay
Problem
The term printed on an invoice is a statement of intent; the real term is the day the money arrived. The reason revenue broke records while the till stayed empty was hiding in that gap.
- C#
- MSSQL
- FIFO matching
- WinForms
Measured
- 250.000+
- matched records
- 1.100+
- customers in one engine
- 178
- weekly periods since January 2023
- Artificial Intelligence Public institution · management information system
Decision support you ask in plain language
Problem
Management did not want a report, it wanted an answer. The analyst layer in between took days per question, and by the time an answer arrived the question had changed.
- Blazor Server
- .NET 8
- Oracle
- MSSQL
- Windows Auth
Measured
- 3
- separate data sources, one question
- 0
- raw rows sent to the model
- 3
- stages of authentication
- Finance Wholesale distribution
Catching the customer who quietly disappears
Problem
A customer does not slam the door; the gap between orders quietly stretches. By the time anyone notices, they have already moved to another supplier.
- C#
- WPF
- SQL Server
- Logo Tiger
Measured
- 4
- risk layers
- 0
- rows written to Logo
- Finance Individual investors
Tools that watch a trader’s own rule across dozens of instruments
Problem
The rule was clear, but watching thirty instruments by eye was impossible; by the time a move was noticed it had already started.
- C#
- WinForms
- Matriks IQ
- DDE
- Live feed
Measured
- 2
- separate markets: crypto and derivatives
- 0
- trading signals produced
- Enterprise Software Food manufacturing
taspigida.com.tr: 100 in all four Lighthouse categories
Problem
The existing site was slow on phones, invisible in search, and impossible to navigate with a screen reader.
- Astro
- Static build
- WebP
- JSON-LD
Measured
- 100
- Lighthouse, in all four categories
- 0
- third-party trackers
Products
Products I have built
-
Fenkron
Carries Logo stock and prices to your online store
Copies the stock and prices of the products you sell online from Logo Tiger into your web store, automatically. If the two stop matching, it reports which product and why. It runs on the company’s own server; nothing outside can connect to it.
-
Price Management System
Turns one price into every dealer price in Logo
You enter one price; it works out the price for every dealer, branch, customer group, campaign and payment method, and writes them all into Logo with one approval. Old prices are kept, not deleted, and every dealer whose price changed receives a PDF list.
-
LoSTaR
Sales and collection reports: who left how much profit
It shows how many days each customer really takes to pay, whose money is most at risk, and how much profit each sales rep leaves behind. Sales figures, each customer’s direction and how sales react to a price change sit on the same screens.
-
OrtaV
Tracks how long customers really take to pay
Pairs every payment with the oldest unpaid invoice and measures how many days a customer really takes to pay. Bigger invoices count for more; last year’s leftover debt, cheques and card payments are included.
-
MüşteKa
Shows you the customer you are about to lose, early
Compares every customer with their own habits: it finds those ordering late, buying less or with a smaller basket, and lists them — most valuable first — with the reason.
-
Corporate Website
A corporate site delivered with its speed and quality measured
A fast corporate site that carries no other companies’ tracking code. Every page gets tidy labels and its own questions and answers, so AI assistants like ChatGPT read it as correctly as Google does.
-
Vocalun
A voice assistant that talks with your site’s visitors
Reads your site, talks with visitors, takes them to the right page and passes their contact details on to you. Added to your site with one line of code.
Questions
Questions about this page
Which sectors have delivered systems?
Public sector (grant management), building-materials e-commerce, poultry and red meat wholesale distribution, wholesale food distribution, wholesale distribution, food manufacturing and individual investors. Each entry states the problem, the work done and the measured result separately; client names are not published without permission, the sector is.
How are the results of delivered systems measured?
Every job carries a performance figure or a quantity: e-commerce synchronisation across 1,551 products, a price management system in use for 15 years, zero raw rows sent to a language model, a maturity engine processing over 250,000 matched records across more than 1,100 customers, a corporate site scoring 100 in all four Lighthouse categories. These are measured outcomes, not delivery promises.