Work
Projects and references. Some clients named, some kept anonymous for confidentiality.
Manag.Web is a business-management platform I build and maintain as my own product: invoicing, double-entry accounting, inventory and fixed assets, subscription billing, and Romanian fiscal compliance (e-Factura, SAF-T). It runs in production at companies including Floricon Salub (waste management), Clodinstal and Utilis.
Dozens of modules on a shared framework, with SSRS reporting - it is the system most of my recent work sits on.
The product documentation is public at manag.zbconsulting.eu.
A modern, responsive management dashboard that surfaces financial health, profitability, revenue trends, receivables aging and operational alerts from a company's own historical data.
Currently in development. The estimation approach has been validated across three different business models with no schema changes. Kept generic while it is pre-launch.
Built the server-side backend and REST services for Bitpool - the platform that ingests time-stamped device data and serves it over an API. Today Bitpool is a commercial IoT and energy-data product (bitpool.com); I built the early backend it grew from.
The work was a full C# solution - server, data model, an uploader with REST services, monitoring, installers, and the certificate plumbing - as backend lead at RISE Technology Solutions in Australia. At the same studio I also built the REST backend for a second product, miSafe.
Built CLIA, an insurance claim-management application, for Venuiti Solutions - claims, financial reserves, recovery, and the reporting around them. C# on Visual WebGui, with a Krypton-based ribbon UI.
I worked as part of Venuiti's team, which built on a Visual WebGui component architecture and included former Microsoft engineers working to a disciplined enterprise process.
Built and extended SIS DEM (the Species Information Service Data Entry Module) for IUCN, the World Conservation Union - the application scientists around the world use to enter and validate the data behind species assessments for the IUCN Red List of Threatened Species.
The work was about a reliable data-entry layer over a shared scientific dataset, used by a distributed, largely non-technical research community, where correctness and clarity matter more than anything clever.
Developed and enhanced the Turtle Research and Monitoring Database System (TREDS) for the Secretariat of the Pacific Regional Environment Programme (SPREP) and the Marine Research Foundation - the system researchers across the Pacific use to record tagging, nesting, strandings and beach surveys, and to track individual turtles over time.
The work turned an existing Microsoft Access database into a proper layered application: a web front end with separate business-logic and data-access layers, plus a desktop tool, packaged so conservation programmes in different countries could install and run it themselves. Built in C# / ASP.NET over the Access data.
Developed Jeeves CRM-ILM, a CRM and internet-lead-management system for car dealerships, for DealerSoft - a Chicago software company. I was the developer; the product was theirs.
Built in Visual Basic on a reusable framework I carried across projects, it handled what a dealership runs on: lead management, deal worksheets, credit applications, and the dealer-specific printed forms each franchise required. It was sold to a number of dealerships over a collaboration that ran for several years.
EZBid is a damage-appraisal and repair-estimating application for dTailTech, a Texas company. It is used by appraisers across the southern United States who assess property damaged by natural disasters - storms, hurricanes, hail - and produce the itemized repair estimates that go into insurance claims.
The starting point was an older DOS program with several linked databases and years of built-up logic. I rebuilt it as a Windows application, keeping the data and the workflow the appraisers depended on, and sourcing third-party components where the job called for them. It shipped with very few bugs and was sold as a commercial product for years afterward.
In hindsight, it is the same work I still do: take a system a business already runs on and move it to a current platform without losing what made it work.
Built and ran a multi-tier business application for EconSeeds, a Netherlands seed company. A layered C# system - separate business-logic, data-access, security, configuration and localization layers, a Visual WebGui web front end and a desktop client, over a dOOdads (MyGeneration) data layer.
I also hosted it and supported the client directly for several years, including documentation and handover so their own developer could take over day-to-day work.
Built zwt, a reusable Visual WebGui application framework - the layered foundation (data access, security, localization, theming and UI components) that line-of-business web apps sit on, instead of rebuilding the same plumbing for each one.
Several of my own applications were built on it, and other software houses built their products on it too - including AboutSoft and BugFree in Italy - with ongoing modules and versioned releases from me.
Two more applications on the same Visual WebGui foundation: AssurantieService, an insurance-services platform with its own theming and error handling and web, desktop and Windows front ends; and DSG, a layered line-of-business web application. Each carries the full business-logic, data-access, security and localization stack.
Both were delivered for Dutch clients, alongside EconSeeds, as part of a portfolio of Visual WebGui business applications built on a shared framework.
On a high-volume permitting platform, fee calculation ran for 30-90 seconds on large permit batches, which blocked the UI and the batch pipeline.
Going through every caller of the calculation engine showed a recursive double-execution and per-row database writes that accounted for almost all of the time. The fixes were to remove the recursion, batch the writes with table-valued parameters, and rewrite the correlated subqueries in the stored procedures. Each change was checked against real production data before rollout; the heavy-batch path dropped by 89-99%.