Background
Sergei Anikin. Based in Tallinn, Estonia. MSc in Computer Science from Tallinn Technical University.
Swedbank (8 years) — internet banking, core system migrations, led 50+ engineers.
Skype / Microsoft (8 years) — senior architect, led integration of Skype into Facebook chat. Engineering manager at Microsoft after acquisition, integrating Skype for Windows 8.
Pipedrive (8 years, CTO) — joined as VP of Engineering, employee #20, scaled engineering from 10 to 400+, grew ARR from $1M to $100M, exit to Vista Equity Partners. Named Top 25 Software CTOs of 2022.
Tuum (interim CEO, core banking). Bolt (Director of Engineering).
Angel investor and mentor in the Estonian startup ecosystem.
How I think
My default is analytical. I classify, simplify, and narrow the option space to find the most viable path — not the most popular one.
I value clarity over comfort. If a project scope feels vague, I'll push until it's concrete. If I think an idea won't work, I'll say so early rather than let it fail quietly later.
I move fast with incomplete information. But not with implicit goals. Before we start anything, we agree on what success looks like and how we'll measure it.
How to propose a project
I take on a small number of projects per quarter. Each one is selected carefully.
The fastest way to start a conversation is a short message with four things:
- —The problem. What's broken, slow, or expensive? One or two sentences.
- —How it works today. The current workflow, even roughly.
- —What you've tried. Even if it's nothing — say so.
- —What good looks like. How would we know it worked?
I'm not looking for polished proposals. I'm looking for a real problem that someone actually cares about solving. "We waste 15 hours a week on X" is more interesting to me than "We want to explore AI opportunities."
How I communicate
I score high on empathy but low on politeness. This is a specific combination worth understanding.
Directness is respect. I prioritize truth and clarity above social niceties. I can be blunt. This is not a lack of caring — it's my way of being efficient and honest. If I challenge your idea directly, I'm taking it seriously.
I sometimes go quiet. I process information internally, especially when the problem is complex or the stakes are high. This isn't disengagement — I'm running through possibilities before responding. If you're unsure whether my silence means "thinking" or "disagreeing," just ask. I'll never be offended by the question.
I'll tell you what I actually think. If I see a risk, I'll name it. If I disagree with the direction, I'll say so. If something isn't working, I won't wait for it to become obvious. In return, I appreciate the same from you.
How projects run
Kodulabor projects are short and focused — days to weeks, not months.
We agree on scope upfront. A clear start, a defined outcome, a bounded timeline. If the scope needs to change, we discuss it openly — not through gradual assumption.
I do the work myself. I'm not managing a team or delegating. When you work with Kodulabor, you're working directly with me. This keeps things fast and avoids the telephone game.
I measure everything. Time spent, tools used, what worked, what broke. Every project gets a structured assessment. This isn't bureaucracy — it's how we know whether AI actually helped or just felt impressive.
If something isn't working, I'll say so. I'd rather stop and redirect than push through on momentum. Sometimes the honest answer is "this isn't a good fit for AI" — and that's a valuable finding too.
What to expect from me
Speed. I work in short, intense cycles. You'll see results in days, not months.
Honesty. I'll tell you exactly what I think — about the project, the approach, and the results. No sugarcoating, no hidden agendas.
Measurement. You'll get numbers, not impressions. How much time was saved, what the quality looked like, where things broke. The full picture.
A published case study. If we work together, the results become part of Kodulabor's public research. You get a working solution. The field gets another data point. Names can be anonymized if needed.
A clear "no" when it's a no. If your project isn't right for Kodulabor, I'll tell you directly and explain why. No vague delays, no ghosting.