A test bench another engineer can use
Vasiliy’s first engineering problem was getting a television signal. New channels had appeared, his antenna could barely pick them up, and he wanted to watch cartoons. Years later, at a research institute, he found himself following a device all the way from an idea to the factory floor. A prototype that worked on its designer’s desk still had a long way to go. Someone on the production line needed a reliable way to check every board that followed. Designing that test interested him as much as building the device.
That experience proved surprisingly portable. On his first software testing assignment, Vasiliy had to figure out network protocols he had never used. Without a specification to consult, he captured traffic and built a small network from the computers he could find. Before long he was explaining the product to the customer in the US. He credits his manager with spotting the responsibilities he was already picking up and helping him take on the next one. It became a model for how he would support engineers on his own teams.
Earning the right to work with the customer
At Deutsche Bank, investment banking projects usually had managers in financial centres close to the business. Engineers in Moscow had a reputation for handling difficult implementations. To manage a project themselves, they needed direct access to the people who needed the work done. A manager gave Vasiliy a chance to test that arrangement in market connectivity. He took the technical project manager role alongside a project manager in London.
Six months in, the customer could get answers directly from Vasiliy and management could ask him for the status of the work at any time. The separate London role had become redundant. The payoff came when a programme spanning several projects lost its manager. Vasiliy put his hand up. By then he could spend his energy on the programme itself. Colleagues already had first-hand experience of someone in Moscow taking responsibility for delivery. Rank still mattered at the bank, but on this particular question he had earned their confidence.
What travels with a client portfolio
Soon after Vasiliy took charge of Russian engineering across the equities business, Deutsche Bank announced it was closing that business. BNP Paribas was interested in serving its clients. Moving the portfolios alone would give those clients a reason to leave if they lost the service they relied on. The negotiations therefore extended to the platforms and the engineers who knew how to run and develop them.
Vasiliy had different cases to make to different people. The buyer needed to understand which systems it would need and why it made sense to retain engineering capacity in Russia. His colleagues at Deutsche Bank were reluctant to part with developers they had spent years recruiting and training. The engineers themselves faced a move to an organisation that barely existed locally. Vasiliy helped establish a BNP Paribas technology centre with offices in Moscow and St Petersburg and put together the systems handover. Getting the conditions right for people to make the move was, in his account, the hardest part.
What a missed deadline means on a production line
At both banks, many of the people who set the rules were effectively out of reach. Chestny Znak offered Vasiliy a chance to help set strategy himself. The national product tracking system was live for its first two product categories. Engineers were still stabilising it, while the next mandatory launches were only months away. A missed engineering deadline would have consequences for the businesses required to use the service.
Vasiliy follows a package from production to checkout to explain those consequences. A dairy needs to reconcile its packaged output with the volume of milk it received. The milk is already in a tank and can spoil while the service is down. At checkout, the challenge is to establish whether this particular package can be sold while the customer waits at the till. Even the round trip from Vladivostok to Moscow would take up a sizeable part of that time. Some of the infrastructure that answers those queries consequently runs closer to the shops.
Planning together and releasing separately
Behind a scan lies a chain of ownership. Packages go into boxes and boxes onto pallets. A shipment can change hands and a correction to its paperwork can arrive later. The system has to establish who owned the goods at the time and which individual codes the shipment contained. Adding a product category touches several points on this journey, right down to the app that tells a shopper what they just scanned.
Business owners and engineering teams agree on a month’s work together. Customers inside the organisation know when to bring a request forward and can settle competing priorities while the teams they need still have capacity. I compare it to renaming Tinkoff as T-Bank. A change that looks like a single event to the customer requires many teams to agree on how it will happen. At Chestny Znak, however, asking everyone to deploy together used to create a queue at integration testing. Separate releases and shared feature flags let teams deliver their pieces independently and activate the feature when the whole chain is ready.
What engineering can learn in the boardroom
A seat on the board gives Vasiliy a view of the work likely to reach engineering two or three years from now. Adding product categories had become a familiar sequence of changes. Conversations with business colleagues suggested that requests from regulators would increasingly require changes deep in the core. Knowing that ahead of time matters for an organisational decision. Someone needs the authority to protect the architecture when several individually reasonable requests reach it at once.
Introducing AI into engineering raises a similar question about who gets to decide. Developers have to work with models hosted inside the organisation and naturally compare them with tools available elsewhere. Examples from teams changing real services helped establish where the internal tools could help. The organisation chose OpenSpec to develop specifications and is adapting OpenCode to its development process. A common foundation still leaves room for a team to write evaluations for its own service. That gives engineers a concrete way to judge an update by the tasks it can actually handle.
When I ask which principles survived all these changes, Vasiliy comes back to volunteering for a problem that has no owner and trusting the people on his team. Even with agents doing part of the implementation, people still have to discover what a user needs and check whether the result delivers it. He wants to give engineers the same room his first manager gave him, with support available when they need it.
What to take away
- 01Direct customer access helped the Moscow team demonstrate it could own an entire project.
- 02A platform transfer depends on retaining the people who know how to deliver the service customers expect.
- 03Shared planning exposes dependencies while feature flags let teams release on separate schedules.