The Oxford mathematician who keeps building companies in categories nobody's quite named yet. His latest is an AI coach that has already taken two million calls.
Enki, the company Kirill Makharinsky runs from a building in San Francisco, has a strange pitch. Its product is described, in plain English on his own LinkedIn, as “human-like, on-demand, and infinitely patient.” Three adjectives that do not normally describe enterprise software.
The pitch is also doing rather well. By the company's own count, Enki has now coached more than two million professionals through the bewildering little forest of tools that modern white-collar work demands: SQL, Python, Looker, Power BI, Tableau, Excel, the rest of it. Companies pay Enki to teach the muscle their teams should have but don't.
Makharinsky did not arrive at coaching by accident. He has spent twenty years inside the question of how grown-ups actually learn. He has co-founded an online travel company that grosses nine figures, a strategy-intelligence platform now folded into NetBase Quid, and a student-entrepreneur network out of Oxford. Three companies, three categories, one through-line: take a process humans do badly, and put a quieter, more patient version of it inside a piece of software.
Enki AI, launched in 2024, is the cleanest version of that thesis yet. It is the company's bet that one-to-one coaching does not have to be a luxury good. That, given enough context about a learner and a company's stack, a model can do the work that used to belong to a senior analyst with a free afternoon and a teaching streak. It is, in a sense, what Makharinsky has been building toward since he wrote his first product spec at Slide.
Which is to say: he is not in a hurry, and he is not pivoting. He is finishing.
“Enki AI is human-like, on-demand and infinitely patient - and tailored to what you're learning, your profile, and your company's tooling.” - Kirill Makharinsky, on launching Enki AI
He was born in Russia and educated at Eton, which is roughly the equivalent of being born in Detroit and educated at Versailles. From Eton he went to Oxford to read mathematics, and from Oxford he came out at the top of the country: an MMath in applied maths with a double first, ranked #1 graduate in the United Kingdom for his year. That is the sort of detail that sits on a CV and looks like a typo until you double-check it.
He spent the rest of his Oxford years running Oxford Entrepreneurs, the student society he became president of in 2004, then turned it into a launchpad. Out of that orbit came YouNoodle - a sort of dating site for startups and the people who wanted to find them - and a habit of being in rooms where someone was about to fund something.
Silicon Valley followed. He landed at Slide, Max Levchin's social-applications company, as a product analytics manager. The timing was excellent. In 2010, Google bought Slide for $182 million. By then Makharinsky had already taken what he wanted from the job: a working understanding of how to read what users actually do, as distinct from what they tell you they do. He has been pointing software at that gap ever since.
Quid, co-founded around 2010, was the next move. The product looked, from a distance, like a search engine for strategy teams - an intelligence platform that mapped industries, ideas, and competitors visually. It became part of NetBase Quid and crossed $50 million in annual revenue. In parallel, almost ridiculously, he co-founded Ostrovok, the Russian online travel agency that would become Emerging Travel Group, parent of RateHawk and ZenHotels, with revenues over $150 million a year. Two categories, two companies, one founder, mostly the same week.
Co-founder. The Ostrovok / RateHawk / ZenHotels family. A profitable OTA doing nine figures of annual revenue. The company that proved he could ship operations as well as code.
Co-founder. Intelligence software for strategy teams - now NetBase Quid. The company that turned messy public data into something a McKinsey partner could actually read on a Tuesday.
Co-founder & CEO. The AI work-skills coach. Two million learners and a thesis: every knowledge worker deserves a tutor that remembers them.
One of the cleaner pieces of internet real estate held by any operating founder. It is just his first name on X. He uses it sparingly.
That is how he frames Enki. The claim sounds large until you notice the company has been quietly stacking learners for years.
Oxford gives out a lot of firsts. It does not give out a lot of #1-in-the-country distinctions. He picked one up in mathematics.
Slide produced more operators per square foot than almost any Silicon Valley acquisition of its era. He is part of that diaspora.
He has been writing small checks for fifteen years. TokBox, among others. The thesis seems to be: pay attention to infrastructure people will need later.
He keeps one. Forty-something photos. The kind of detail that betrays a founder old enough to remember when the internet had hobbies.
The Enki argument runs roughly like this. Companies spend serious money on tools - Looker, Snowflake, Mixpanel, Tableau, Power BI, the rest - and then leave employees alone with them. Tool adoption is the silent loss line on every IT budget. Workshops do not solve it; documentation does not solve it; an annual training week does not solve it. A patient tutor, available in the moment of confusion, does.
What is new is that the patient tutor no longer has to be human. Enki AI is the productization of that idea: a coach that knows what your company actually uses, what you are trying to do this afternoon, and how far you got the last time. The Enki team has spent years building the structured content and role-specific exercises that make this kind of personalization plausible. The AI sits on top of that scaffolding, not in place of it.
Makharinsky is, by training, an applied mathematician. The applied part shows. He does not talk about education as a calling. He talks about it as a system that is badly tuned and worth re-tuning, and about Enki as the company that finally has the data and the model class to do it.
His critics, such as they are, would point out that the AI-coaching field is now crowded. His answer is the boring, correct one: distribution and depth. Two million learners and a back catalog of curriculum is not nothing. It is, in fact, the moat.