Thinking Out Loud: The AI Hype Continues 🤔 (Post in Russian can be found above)

Hey everyone! Today is my birthday, and I like using this day to reflect and set personal goals for the year ahead.
Since this channel is about career, I wanted to share some thoughts on what I plan to focus on next.

New AI Chapter 🚀

Right now, I'm working on client engineering for AI tooling, which means diving into an entirely new domain for me. For the first time in almost two years at Meta, I'm genuinely excited about what I'm working on. Building the UI layer for AI tools isn't particularly difficult. The real challenge lies in optimizing how models operate across both the client and server sides. A well-designed client can turn a relatively simple model like Gemini Flash into a surprisingly powerful tool. I believe that in 2026–2027, large companies will increasingly invest in developing their own LLM harnesses. It's not enough to simply own the model—you also need to control its output. Client-side optimizations can reduce context consumption by 2–3x. When you scale those savings across an entire company, considering the cost of tokens, you're potentially saving tens or even hundreds of millions of dollars.

What's Changed for Me❓

I now report directly to a VP and a Director, and I work closely with IC7+ engineers from the Dev AI Infrastructure team. And honestly, it's challenging! 😁 The concentration of exceptionally talented rock stars is incredible. There's a lot to learn. At the same time, I think I'm in a strong position to pursue an IC7 promotion. If not now, then when?

Is It Worth Going for IC7❓

I think so. Total compensation for an IC7 typically ranges from $2M to $4M, depending on the equity package. On top of that, if you consistently meet expectations, your compensation can grow by roughly $300K–$400K per year. More importantly, this is the level where you start having significant impact on the company and gain opportunities to work directly with VPs and Directors.

What's Next for the Channel❓

I'm joining the hype train a little late, but over the coming months I'll try to write more about client-side AI engineering and model optimization. That's what I'm actively learning and working on right now. I'll also try to share insights about the path to IC7 and what it actually takes to get there. As usual, posts will probably be infrequent—but hopefully worthwhile.

Question for You ❓

What would you like to see more of on this channel? I'd love to publish more content, but realistically I don't have enough time to write frequently. Livestreams are much easier for me at the moment. Would there be value in sharing research papers, architecture documents, books, and other resources related to LLM clients and AI agents?

Thanks for reading! 🙏

YouTube 📹 — Medium 📑 — LinkedIn 👨‍💻