Every platform has problems the team learned to live with.

Adobe Experience Platform. Teams own their services end to end: code, instrumentation, production operation and on-call.

Edge Ingestion, since 09/2025. Backend and reliability engineering on a core Edge Network service. Split the service's operational and analytical metrics into two lanes, so that alerting and autoscaling run on a lean, reliable signal and the analytical detail lives apart from it: more reliable operational monitoring and room to scale. Owned the end-to-end test framework used for release validation and defined and instrumented the ingestion SLIs and SLOs; delivered distributed tracing with OpenTelemetry across the request path, with sampling controlled at runtime.

System Observability, 06/2023 to 08/2025. Owned the cost and reliability of the Prometheus-based monitoring stack used by engineering teams across Adobe Experience Platform and other Adobe organisations, a fleet of more than a thousand Prometheus instances; hands-on in configuration, CI/CD and code (Java, Go, Python). Analysed what the fleet's tenants actually used and set safe minimum CPU, memory and disk defaults from it, with per-namespace overrides for heavier workloads; the defaults apply automatically to every new tenant. Designed horizontal and vertical Prometheus scaling and built the auto-configuration layer that applies it. Availability from 99.5% to 99.99%, through alert filtering and prioritisation, self-healing automation, error budgets, restricted change periods and wave deployments; led postmortems. Wrote the alerting and monitoring guidelines most of the tenant teams adopted on recommendation, through reviews, office hours, onboarding and training across the organisation; designed the organisation-wide move to a new alert-routing platform as one configuration line per team.

A Series-B platform on which banks and insurers build and launch financial products; clients included BRD, Banca Transilvania, Erste, Vodafone and Vienna Insurance Group. Initiated and built the NuGet package build system the company adopted: templates, versioning and breaking-change rules, and dependency pipelines that rebuilt and retested only the affected packages, wired into GitHub Actions. Introduced the SLI/SLO measurement criteria and the bottleneck-discovery and profiling method (Application Insights) the performance team worked by. Tech Owner after six months: technical lead for five backend developers, whose code-review standards I set, and the coordination point for ~40 people across business and technical teams; technical driver of a product letting non-technical users automate core-banking processes (Flowable BPMN engine).

Global intellectual-property services group; DIAMS iQ, its enterprise .NET platform, for clients including Nestlé, Coca-Cola and Volkswagen. Owned the design and build of the trademark-renewal automation inside the existing monolith, after a multi-year greenfield rewrite had failed: ~10,000 person-days a year of manual work removed, and the people it freed moved into new business areas. Redesigned the delivery flow end to end, build, tests and deployment, from a working day to under an hour, the build step from four hours to twelve minutes; it became the organisation's default path. Led a development team on the renewal automation and, from 2020, served on the Engineering Steering Committee, the deciding body for technical direction, KPIs and engineering standards across the ~50-person software department over four countries; introduced performance testing (k6, from zero, run daily) and CI/CD validation gates for the whole department, proposed by me and adopted through the committee. Built the Grafana dashboards that directors and VPs used for oversight and engineers used during incidents.

Enterprise SaaS for event and entertainment payments: one shared .NET platform on AWS, clients including Apple, Dell and PwC. One of four architects; owned code-level performance across a ~40-person engineering organisation. 10x faster critical paths while the client base doubled and data volume grew 5x, through asynchronous communication, eventual consistency for non-critical paths, in-memory caching and materialised views. Set the performance rules every implementation passed through in review, and built the IQueryable guardrail that flagged unfiltered database calls before code review and became the team standard.

Built the iOS framework and application for an encrypted mailbox for confidential medical documents, the system that became Germany's electronic patient record (ePA).

Took over an existing C++ and C# codebase on the Acquity liquid-chromatography line and led and trained the new Brașov team extending the US team: introduced Scrum and prototyped a UI test-automation suite.

Multi-screen TV technology for pay-TV operators; clients included Swisscom, Kabel Deutschland, Siemens, Roku and Vestel; the framework work behind 3Ready, 3SS's product platform. Microsoft Mediaroom for Swisscom (500K daily active users, ~50K concurrent at peak): C#/.NET backend performance engineering and client-channel optimisation (caching, local storage, prefetching) that cut server load and raised perceived client speed. Built an iOS framework for controlling multiple TVs from one device over GPS, WiFi and ultrasound, reverse-engineered from vendor behaviour when the documentation failed; native iOS and Android clients.

Siemens (healthcare IT), MoreTV Broadcasting (digital TV), Testmetrix (semiconductor test equipment), Decsoft (web and CAD software). At Decsoft, a ~10-person web and CAD software company, ran the development department as its Software Architect, a first leadership role with the hiring calls and the company's first version control and deployment processes; at MoreTV, the middleware layer between the user-interface code and the low-level drivers, a clean API boundary that was also the working bridge between the two teams.

Took Carmira from idea to a live product with ~100,000 monthly active users at peak (2021): owned the product, its roadmap and its economics, and ran production operations. Cold-called ~200 dealerships and converted 20 into paying customers; rebuilt the marketplace around their feedback, adding the dealer CRM they asked for. Led a team of four, one engineer grown from intern to technical lead; drew a majority-stake investment offer from Romania's second-largest car-dealer group (2021).

Architecture, delivery, operations, hiring, sales and P&L; team of up to four. Parfumdreams, Douglas Group (2023): the modernisation of its multi-country e-commerce operation, storefront through warehouse, stock and courier systems, onto .NET Core microservices on Kubernetes. E-scooter rental platform (2018): end-to-end system design for a local investor, from the code to the hardware selection and its custom integration, with the vendor negotiations. Gaming-operator prize-flow audit (2017): for a Romanian land-based operator with thousands of locations, my observability and audit tooling found ~€20,000/day (~€7M/year) of untracked losses.