Staff Engineer @ Xsolla · Frontend Craftsman · Web3/Ethereum Architect · AI Systems
Deep in the pixels, deep in the systems behind them.

Every pixel and every system behind it should hold up under scrutiny.

I'm a Staff Engineer at Xsolla, and first and foremost a frontend engineer — someone who still hand-tunes layout, motion, and rendering, and treats the interface itself as a craft, not a checkbox. That same rigor carries into everything around it: architecture, Web3 integration, and the AI systems I now design and operate in production.

• 🎨 Frontend Craft: hand-built SVG/animation, design systems, SSR / ISR / CSR / SSG, reactive state machines
• 🧩 Architecture at Scale: modular SDKs, event-driven bridges, strict typing, monorepo tooling
• 🔐 Authentication & Security: Passkeys (WebAuthn/FIDO2), PKCE, OAuth 2.1, JWT flows, signature validation
• ⚙️ Web3 Integration: Ethereum smart contracts, Wagmi / Viem / Ethers / Web3.js
• 🤖 AI Systems: LLM/agent pipelines, RAG, MCP tooling — designed and run in production, not just prototyped

Skill without rigor is just a demo. This is how the work actually ships.

• 🔍 Nothing merges unreviewed — including my own code; review is where quality is actually enforced
• 🧪 Tests as a design tool — unit → integration → E2E, written to pin down behavior, not to pad a coverage number
• ♿ Accessibility and performance are budgets, tracked in CI, not a "nice to have" for later
• 🧭 Staff-level ownership — I set technical direction across teams, mentor engineers, and stay hands-on through incidents, not just architecture diagrams
• 🤝 Engaged, not just present — code review, design review, on-call, and pairing are where I actually spend the effort

I don't just call an API and ship it — I design the whole circuit: context, guardrails, tools, and the feedback loop around the model.

• Agent & LLM Pipelines — prompt/context design, tool-calling, multi-step agent orchestration, guarded outputs
• RAG & Retrieval Systems — source curation, chunking/embedding strategy, relevance evaluation
• MCP & Tool Integration — wiring LLMs into internal APIs, SDKs, and developer tooling
• AI-Augmented Developer Workflows — CI copilots, review automation, internal tooling built on top of LLMs

flowchart LR
  A[Product Need] --> B[Prompt & Context Design]
  B --> C[LLM / Agent Orchestration]
  C --> D{Tools or RAG needed?}
  D -->|yes| E[Tool Calls & Retrieval]
  E --> C
  D -->|no| F[Guarded Output]
  F --> G[Production Integration]
  G -.observability & evals.-> B

"The code must explain itself — and so must the model."

• Composition > Inheritance
• Type safety mindset (TypeScript strict / Rust ownership)
• Clean Architecture · SOLID · CQRS
• AI is a system to be engineered, not a black box to be trusted blindly
• Performance is a feature, accessibility is a requirement

Exploring the edge between AI, technology, art, and structure.
iShavlovsky ⚙️ 2026