←TripleTen

TripleTen is a personalized learning service that takes you from gaining skills to employment. Named by Fortune as the best overall tech bootcamp in the US, it has over 20k monthly active learners across the US and Latin America. TripleTen is a part of Nebius, a $20B NASDAQ-listed AI cloud company. This is also the funniest place I’ve ever worked.

Platform

HomeLearning pathLessonQuizProject

The product was born in the early 2020s—a time of economic uncertainty, especially in the US. For many, moving into tech became a path to stability and a better life. TripleTen became the best way to learn a tech job for people with busy lives, thanks to its mix of full control over the student experience, strong learning design, and deep expertise in tech and AI.

I joined TripleTen as part of the founding team and spent four years shaping the product design and adapting it to the needs of demanding US students. Over time, I took on broader design leadership across the organization—supporting other designers while keeping my sleeves firmly rolled up.

Dot, your AI buddy

One of the biggest challenges in online learning is the lack of real-time support. In a classroom or even on Zoom, it’s easy to ask a question and get help right away. But when you’re studying online, that instant feedback is missing. That’s why TripleTen made human support—tutors, learning coaches, and community—one of its main priorities.

The problem is that support is often either fast but shallow (like a quick reply from a support agent) or deep but slow (like a tutor’s detailed explanation). And keeping depth and quality consistent is especially hard at scale.

AI changed that. In 2025, with the rise of reasoning models and agent frameworks, smart and responsive AI tutors suddenly became possible. That’s when Dot was born.

Dot started as a quick experiment on our Discord server and stayed there for almost a year. That time showed us how much value students got from it, so we decided to turn it into a core part of the learning platform.

We began where students struggled the most—coding projects. It was also where we expected the most feedback. That’s why the first chat we built lived inside the VS Code extension.

From the start, we took a “ship to learn” approach—built the simplest possible chat, added many events and feedback loops, and started watching what students asked, when, and how. A huge part of this was powered by PostHog’s toolkit for surveys, analytics, session replays, and feature flags, which we used heavily.

After the launch, we saw heavy usage of the extension, confirming our hypothesis: this kind of help was genuinely valuable, and the entry point worked. Our next step was to identify other key moments where students needed support most and make Dot easy to access there.

Over the next six months, together with a small team of engineers and researchers, we integrated Dot into the platform and iterated on its conversational experience, tools, memory, voice mode, and various entry points across the learning journey.

We managed to make Dot more than just another AI bot—it became a real buddy on the platform that helps right when you need it. Our A/B test clearly showed the impact: Dot’s support increased project submissions by 34%. It quickly became the platform’s most-used feature, with 47% MAU and over 200k messages sent each month. Equally important, Dot’s lively personality and tone of voice brought widespread student love, tons of positive feedback, and a 91% satisfaction rate.

Projects on the platform

TripleTen’s approach is built around project-based learning—and projects are the biggest, most complex part of it. Students spend most of their time working on them, and that’s also where they need extra support.

In the past, these projects relied on a mix of external tools, while the platform itself was mostly a textbook with instructions. Once you started a project, you were on your own, setting up a complex coding environment locally.

That created a painful UX barrier for students and a blind spot for the product team, since we had no visibility into the part where students spent most of their time.

So we decided to bring projects onto the platform itself. Today, most professional tools students use (VS Code, JupyterLab, Google Sheets, Figma, and more) are web-based, which made it technically possible.

It was a big UX challenge that required rethinking the whole platform layout so students could work comfortably in a complex environment while still having easy access to all the platform tools. The goal was to make it feel like the same familiar place—just with the features of a full professional workspace.

Another major challenge was simplifying project statuses. Project workflows could get really complex, so I had to break them down into the clearest possible model. Now each project requires completing only a few steps that make immediate sense to students. At the same time, the layout and look of each step can vary significantly depending on the project type and status.

It was a fun design project but also a big technical one—especially when it came to making the platform and external workspaces work seamlessly together. With a tight-knit team of engineers, we built the first VS Code version in three months, but the full rollout, iterations, and support for more tools took another year.

Projects

VS CodeSpreadsheetJupyter Notebook

As a result, project work became a native part of the platform. Students can now work on projects without the hassle of setting anything up locally, and they have access to all familiar platform tools. The team has full visibility into what happens during projects, making things like the A/B experiment with Dot support mentioned above possible. The entire learning process now happens within the platform, and students spend 31% of their time in the new project workspace.

Observability

In my last year at TripleTen, one of the big things I focused on was helping us understand how students actually study on the platform.

Before that, our product analytics setup was quite limited, and we had almost no way to observe real student behavior. We often didn’t know whether what we shipped actually worked. UI decisions lacked context, and we mostly relied on personal opinions.

When I started researching analytics tools, one clearly stood out—PostHog. Beyond its hog-themed aesthetic, it proved to be a powerful and flexible observability toolkit.

I set up and maintained the PostHog integration from early experiments to a full rollout across all platform surfaces. It was an interesting challenge because it combined setting up structure inside PostHog, maintaining integration code, resolving issues, and driving awareness and adoption across the company.

Over time, PostHog became the core analytics and experimentation platform for the product team and was broadly adopted across the org. Here’s what we achieved with it:

• Iterated on freshly released features based on usage data, session replays, and survey feedback.
• Ran experiments to measure impact—from getting 72% more feedback responses by tweaking a button label to proving that 34% more students submit their projects with Dot support.
• Started rolling out new features gradually using feature flags.
• Unshipped several low-usage features based on insights from automatically captured data.
• Moved student activity tracking for the learning support team from a clunky in-house system to reliable PostHog events within the main infrastructure.

Visual language

One of the things I truly loved at TripleTen was shaping its visual language and bringing the brand’s voice into the interface. During my time there, I was the key maintainer of our shared design system.

One of the hardest challenges was keeping everything consistent and helping the other 25 designers across product, brand, and curriculum teams align and contribute effectively to the shared system. To make it easier for everyone, the system was structured as a tree of modules—from broad foundations to more specific UI libraries, brand components, and curriculum guidelines. These Figma files and the shared approach behind them took shape through regular design critiques and a steady design system routine.

The collaboration between different design functions was crucial in shaping TripleTen’s unique tone of voice. I always wanted learning to feel approachable, supportive, and not boring. That’s why our ToV showed up not only in the copy but also in the small details of the interface.

In my last year, I took this further by stepping into design engineering, shaping not just visuals but also interactions and polish. My code contributions added up to 47 PRs across new features, polish, papercuts, and analytics.

Team

The best part of working at TripleTen was connecting with so many amazing people. TripleTen was always a lot of moving parts: tech folks building the platform, education experts shaping curriculum, support teams, career services, and more.

I started in a small engineering team focused on the platform. Later I took on projects with curriculum, support, and others. I tried to make design a glue that tied the whole experience together and kept us from sliding into Conway’s law. Over four years I went through so many different team setups that it honestly felt like working in several different companies.

Design in TripleTen was also very diverse. We were organized into three pillars: communication, product, and curriculum design. All treating design as a glue, we collaborated closely, shared a visual guideline, design system, and processes. I’m proud I helped this community of bright folks grow—supporting with rituals, tooling, strategy, and hiring.