Slavik Shynkarenko, Head of Engineering Effectiveness at AutoScout24 and Trader

I’m Slavik Shynkarenko. I lead Engineering Effectiveness at AutoScout24 and Trader — a global function, spanning Germany and Canada, with one mission: make the engineering investment of 800+ engineers transparent, optimised and aligned with business strategy. In practice that is three questions asked on repeat. Where does the cloud and AI money go? Is it spent well? What did it build?

The answers live in my function. FinOps for AWS and AI: cost tracking, budget visibility, optimisation. Engineering CAPEX: attributing engineering effort and AI usage to the work they built, and capitalising it. And the engineering metrics and reporting for the whole group. Cost-efficient AI starts with how teams use it, so I help shape the tools and models they build on.

The function runs AI-native. Two AI teammates I designed and built, Nex and Fyn, do a good share of the legwork: Nex runs sprint health and Jira across three Atlassian sites, Fyn watches the cloud and coding-agent bill and tells each team where its money went. Next comes automation that classifies work for CAPEX and links every AI token to the ticket it served — so reporting runs itself, and engineers spend their time on engineering.

The job is part arithmetic, part architecture, part diplomacy. Mostly it is showing teams what their choices cost — in euros, in tokens, in on-call — and handing them the lever that changes the number. Everybody enjoys a smaller bill.

I came to it from the other side of the table. Until September 2026 I ran Platform Engineering for the same group: four teams on two continents, a multi-million-euro cloud and AI budget, and the paved paths every product team ships on — compute, incident management, observability, infrastructure as code, AI tooling. We consolidated businesses on a single cloud provider, AWS, off on-prem data centres and Azure. Every paved path is a trade: local freedom for global throughput. I have made that trade in both directions and learned where the balance sits. Having run the machine is what makes me good at measuring it.

How I got here

I started in 2008 in Ukraine, writing web-server extensions at Helicon Tech — IIS internals, the kind of code where one bad assumption takes the whole server down with it. Then e-commerce and CRM platforms at 4ib: senior engineer, then CTO by 2014. That promotion is where I learned that being good at the technical part and being good at the leadership part are almost unrelated skills. I learned it the expensive way. By the end of 2016 I was thoroughly burnt out, and I left.

2017 was a gap year in the honest sense — savings, a little consulting, a lot of drifting. It’s also the year I opened an empty repository and started piling up ideas for a programming language, right before I moved to Germany and let the whole thing gather digital dust. Munich came with the guest-messaging platform at TrustYou — hotels, WhatsApp, WeChat, an early AI chatbot — and in 2020 with AutoScout24: reliability and cloud engineering first, then leading it. In January 2025 I took over Platform Engineering for the group. In October 2026 I moved to Engineering Effectiveness — FinOps for AI and cloud, engineering CAPEX and metrics, and the home of the two AI teammates I had built along the way. Somewhere in there: a master’s in computer science from Dnipro University of Technology, an organisational-leadership certificate from Harvard Business School Online, two reorgs, an acquisition, and the particular comedy of running one organisation across two continents and a six-hour time difference.

Most of what I know came from getting it wrong first. That’s mainly what I write about.

2024 and 2025 looked excellent from the outside — bigger title, bigger org — and were hard on the inside. A divorce, an empty new flat, a year spent on autopilot. What pulled me back was building things again for no better reason than wanting them to exist.

By the numbers

  • 800+ engineers in scope, across nine businesses — one group, separate companies, separate engineering teams, two continents.
  • A multi-million-euro cloud and AI budget — first as its owner, now as the function that makes it transparent and efficient for every engineering domain.
  • One cloud provider, AWS. My teams consolidated businesses onto it — some off on-prem data centres, others over from Azure.
  • 54% fewer high-severity incidents after rebuilding incident management around data and learning — and then one incident process for every business in the group.
  • Two AI teammates in production — Nex with about 110 tools across 29 domains, Fyn with about 60 — designed and built by me, owned by my function.
  • 18 years, engineer → CTO → Head of Platform Engineering → Head of Engineering Effectiveness. Two reorgs, one acquisition, one six-hour time difference.

How I run things

  • Let agents do the legwork. Cost tracking, attribution and reporting are work for software. People get the time back for decisions and for engineering.
  • Make the bill visible. People change behaviour when they see what a choice costs, in the place where they make it.
  • Say what the standard costs. Every paved path and every mandate trades local freedom for global throughput. I make that trade out loud, with the price attached.
  • Grow the leaders. Engineering culture compounds faster than any tech stack, and managers and technical leaders are the people who carry it.
  • Stay close to the code. Building keeps my judgement sharp, so at night I build a compiler, a sky catalog, a coding agent and a maths tutor for kids.

What I build at night

Miri is a GPU-first programming language, written in Rust, compiled with Cranelift. You write ordinary code; Miri compiles it to run across thousands of GPU cores at once. The compiler writes the kernels, the buffers and the glue.

It started as that 2017 pipe dream: a language as ridiculously fast as C++ and as joyful to write as Ruby. Then LLMs arrived, GPUs became the default place computation happens, and the idea stopped being nostalgic and started being obvious. I document the whole thing — architecture decisions, type system, the lessons along the way — in the Miri devlog.

SpaceCatalog.org is the other one, and it’s the oldest idea I have. I’ve been obsessed with astronomy since I was four, staring at the craters on the moon and wondering whether someone had hidden Kinder Surprise toys inside them. SpaceCatalog is how I reconnected with that: an open catalog of everything above us — 213,698 stars, planets, galaxies and black holes distilled from 60,091,093 sky sources across sixteen source catalogs, merged into one searchable index with an open API. The first of more space projects.

Niobe is my own open-source coding agent for the terminal. AI coding at full speed, with full awareness — every action the agent takes, every decision and its reason, every file it touches and the running bill, visible as it happens. It went from first commit to public release in under three weeks.

Lernen mit Nelly is the newest, and it began at our kitchen table, practising maths with my daughter. It’s a maths tutor for grades 4 to 7, taught by Nelly, our family’s beagle puppy: 67 topics, a worked example first, then practice, and a clear explanation whenever an answer goes wrong. Free for every family and classroom.

Away from the keyboard

I shoot photographs — concerts, rain, river surfers, whatever the light is doing — and I make Neapolitan pizza with genuinely unreasonable rigour. 24-hour dough, 450 °C, zero compromises.

What you’ll find here

Written from practice: engineering leadership, software and platform engineering, the Miri devlog — and, from here on, the economics of engineering with AI: what it costs, whether it pays, and how to build the agents. Real numbers, named tools, and every post grounded in work I have done myself.

Start here

If you have ten minutes, these four are the shape of it:

Elsewhere

Say hi

If you’re working on AI and cloud cost, on engineering CAPEX, on measuring what engineering delivers, or on agents your teams will love using — or you’re building something ambitious and want to compare notes — write to me: slavik@slavikdev.com . I read everything.