Car Sharing: Getting Operations in Order

Artem was born in Tashkent in 1991 and lived there for over twenty years. He earned his first money in the eighth or ninth grade, around 2004: he bought a domain, put together a one-page site and sold, for 99 cents through WebMoney, an archive of books about getting rich. Sales ran at five to ten a month, which covered the night internet tariff from six in the evening to six in the morning, cigarettes and ice cream. The promise of new books at no extra charge was not a stroke of genius: he simply did not know how to restrict access to the archive technically. Next came a casino assembled from the cut-up scenery of a 2D game: at fifteen or sixteen he wrote to the publisher asking permission to use the illustrations, got a yes within two days, and shut the project down when a player spent one night beating his primitive noughts-and-crosses algorithm.

Then came ten content portals living off contextual advertising, out of which a content exchange grew: the portals closed, the exchange stayed and was producing hundreds of articles a day. He and a colleague got the unit economics wrong by an order of magnitude, pricing, as he now guesses, per character instead of per thousand characters, and only realised it a decade later. He did finish university: he failed his first semester, went through a half-year discrete maths course in a single night, got a three for it, and took away what he calls the main skill of those years, dealing with grown-ups you depend on. In 2013 he moved to Russia, to Tver, and joined a startup doing offline surveys: a web builder plus a tablet app for interviewers, a competitor to SurveyMonkey, acceleration at the Internet Initiatives Development Fund and pitches to investors in front of the team. Mail.ru Group later bought that product, and at CityMobile Artem filled in his own engagement survey through it.

At the publisher Alpina he landed in its small digital unit: a corporate e-library serving two hundred companies, run by a tiny iOS, Android and backend team. Tasks there were prioritised on an Eisenhower matrix and every one was counted in money, and clients were sometimes quoted several costed scenarios — a habit he kept as a CTO, bringing the chief executive three scenarios with their economics and his own recommendation, not one ready decision. The ambition of a SaaS platform for offline chains, where in a bookshop chain's app a paper book unlocked its free electronic twin, ran into an absent market. Then he felt short of high load and moved to a dating service bought by Mamba: ten million profiles and matching built on Sphinx, recomputed nightly on six large machines. The engine rewrite he had joined for never started, so after four and a half months he left, spent two months on Android, and came back to the backend and the hype around microservices.

Then food tech: Mail.ru Group failed to buy Delivery Club and assembled a competitor from scratch, where within a month Artem went from second backend developer to head of the whole backend. Two and a half months of six-day weeks later the news broke that Delivery Club had been bought after all; the team got out the rum, and a candidate who turned up for an interview that evening was honestly invited to join the table. At Foodfox, later Yandex Eda, he built order delivery to tablets in restaurant kitchens, a service that grew into a second monolith and took eight years to break apart. He joined CityMobile on a downgrade, turning down a head-of-development role at another car-sharing company, and a month later the CTO sent him to the taxi service inside VK. In a year the team grew to seven or eight people, and taxi inside VK was doing about a quarter of all rides: 125,000 out of half a million a day in 2020.

CityMobile had no frontend at all, so Artem became its first frontend developer himself: twenty to twenty-five interviews a week, the same question about the fine points of MobX every time, and the documentation opened after each conversation — fifteen developers hired in half a year. Then came sixty people in client applications, one-on-ones with all of them inside ten days, and a platform unit that twice failed to take off under technically strong but non-entrepreneurial leaders. He defended the strategy for a new ordering platform in front of the CTO, and on 11 March 2022 the company was gathered and told it was closing. That evening Lyosha Sergeev, chief executive of CityDrive, called; the three days he took to decide he calls the hardest of his life, with no SWOT analysis in them, only a commitment to give the job a year. On his second day marketing came for permission to send a push campaign: he granted it, and the service went down.

The company he inherited: a flat team of up to thirty, some two hours of downtime a week, three releases a week of which one ended in full downtime and another in server errors, half an hour down on a Friday evening counted as ordinary, and code shipped by git pull onto four virtual machines with the fourth regularly forgotten. Rebuilding operations took eighteen months: Kubernetes, automated releases with canary rollouts and rollbacks, on-call rotations, severity levels, a duty to notify the Moscow transport controller past fifteen minutes of downtime, and IT grown to around a hundred and fifty. GPS jamming, first felt in May 2023, produced a work stream of its own: telemetry from car units and client apps, cell tower and Wi-Fi databases, a predicted coordinate, a search task for a technician, renting by number plate. The DS/ML unit replaced dozens of manual selfie-to-document checkers with a model and three to five assessors. Car sharing, he says, is a hard business with no night maintenance window: behind every incident, he tells engineers, stands someone who popped out for milk and cannot unlock the car in the rain.