This week on Coffee with Calyptus, we sat down with Alexey Sudachen, a systems veteran turned blockchain protocol engineer with deep experience at Kaspersky, Spacemesh, and Analog. Alexey shares how burnout from low-level development led to a rebirth in Web3, and how he’s shaped performant, flexible systems using Rust, Go, and Python.
You’ve gone from deep system work at Kaspersky to develop blockchain protocols and services at Ethereum Classic, Spacemesh and Analog. What sparked your shift toward blockchain protocol engineering, and how has your mindset changed along the way?
In the first place, it was just a wish to do something else. I had been working in system development since the early 2000s: GSM mobile services, VoIP switches, real-time video translation, Virtualization, DLP... This involved a deep dive into system internals, developing not just services but also drivers for a wide range of systems, including UEFI firmware. My main languages were C++, C, Python and Assembler.
I’d like to say - “I realized the landscape of software was shifting, and to stay at the forefront, I made the deliberate decision to pivot away from my established toolkit and immerse myself in modern languages” … No, to be fair I just got tired from C++ and system development.
In 2017, I switched my professional path dramatically, abandoning old tools and starting to use Go and Rust. I only kept Python in my toolset because, yeah, it’s a great tool for automation. I think that in 2017, the only industry where you could find Rust/Go innovative work was blockchain. So, I quickly found a job at Ethereum Classic. Honestly, I didn't do much in that project and left it without a clear understanding of whether I wanted to be part of that world.
After that, I tried working as a Data Scientist for a while. I think I did well but found that it obviously wasn't for me. I prefer to focus on developing systems rather than exploring data. However, it was an excellent experience that gave me a real-world understanding of what Data Scientists and Data Engineers do.
So, I returned to the blockchain industry and started working at Spacemesh. Spacemesh is a very unique project from an architectural perspective. It really expanded my vision of blockchain architecture and the ways we can develop protocols. I started to learn about different approaches to building blockchains, the math behind the scenes, and how to observe and debug such systems. And… I burned out.
I then tried working on projects outside of blockchain but quickly found that I missed the dynamics of a fast-evolving technology and the challenges of decentralization. So, I returned to the Web3 world, equipped with a broader perspective and a clearer understanding of the challenges I'm most passionate about solving.
At Analog I discovered another angle on blockchain services: the customer's perspective, focusing on how they use on-chain data for analytics, decision-making, and automation.
At Analog, you built a blockchain-agnostic indexing backend in Rust. What were the toughest architectural decisions you had to make to keep it both flexible and performant?
In startups, I find the main trade-offs usually aren't in the space of flexibility and performance, but rather how I split development resources among features, testing, and observability. This is especially important for distributed systems that handle financial transactions. I find it's better to decide carefully at every level what's really needed to build a working product with the necessary functionality, a low risk of failure, and decent observability to solve problems as they appear. I also prefer to always expect that MVP and beta releases will have serious issues, even if I haven't found any.
Nevertheless, the toughest decision in any project I've worked on was (and I think always will be) cutting out unnecessary functionality and avoiding over-design while keeping the architecture open.
In rapidly evolving fields like Web3 and AI, how do you determine which ideas have true, lasting potential versus those that are just part of the current hype cycle?
I appreciate how hype can popularize new technology. However, I’m more focused on how the development of Web3 and AI projects is changing the business and technology landscape.
Right now, I see great potential for developing Web3 in two different areas. The first is in low-latency operations on both real-time L1 blockchains and very fast rollups. The second is in Data Availability services, which are normally used by rollups but have potential use-cases beyond just supporting intermediate information, especially for services that actively produce and process large amounts of data.
Regarding AI, it’s already a very useful assistant technology that helps dramatically reduce research time while improving quality and consistency. It also speeds up development, reducing the time-to-market, which allows ideas for new features to be tested faster and more rapidly.
You mentioned using AI assistants to boost your productivity during off-hours R&D. What specific tools or workflows have made the biggest impact, and how do you see AI reshaping how engineers prototype in Web3?
First is Deep Research. There's no doubt that even though I can read and understand complex texts faster than many of my coworkers, an AI does it much faster — 100 times faster, really, and it doesn't get tired. I physically cannot process that much information to find valuable insights so efficiently. AI can also present information in different forms like text, infographics, or audio, which helps a lot when learning something new from the research.
The second is assistance with writing. I’m not a natural writer, and it's very hard for me to produce large, well-written texts. With an AI assistant, it becomes much easier. This applies to everything: research conclusions, idea presentations, documentation, and more.
The third is coding. I don't personally like vibe-coding as a tool, but I'm glad such instruments exist because they allow people who can't code to implement the first version of their ideas or take the first steps in learning. I use a coding assistant in two ways: for auto-completion suggestions and code reviews, and for writing portions of trivial code and tests. For the second part, I treat it like a developer I'm leading, giving it a clear ruleset, requirements, and acceptance criteria. This approach isn't much different from working with junior and mid-level developers. It speeds up my work significantly and, more importantly, saves the emotional energy that gets drained so quickly by simple, boring tasks.
I think AI is already dramatically reshaping the work of software developers, and not just in Web3. Using AI in your work is like using high-level languages and an IDE to develop software. Anyone who doesn't use AI at all will not just be very slow and inefficient, but also limited in their ability to develop much more complex solutions. This isn't just my opinion; it's what is already happening.
Calyptus Intelligence 🤖
Fast Hiring Reviews:
Drop this image + 3 inputs into ChatGPT/Claude:
- Resume
- Job description
- [Optional] Company tone
You’ll get: summary, skills, fit score, red flags + a LinkedIn DM as output
