7.2 KiB
OLEG PROSKURIN
AI Engineer & Tech Lead
usulpro@gmail.com · Telegram @usulpro · LinkedIn · GitHub · Dev.to · Blog
Summary
AI Engineer with 9+ years of production engineering across Frontend, Backend, Serverless, Node.js, and Headless CMS / eCommerce platforms, including 5+ years leading remote projects for US, UK, EU, and AU clients. Currently Tech Lead on PrimeUI (primeui.com), an AI-native website generation platform launched Feb 2026: 3 independent AI flows (chat, page generation, content writing) orchestrating Claude / GPT / Gemini via Mastra, MCP server as integration bridge, RAG-based component selection, and a code-export pipeline that ships production-ready Next.js sites.
Skills
AI Engineering
- LLM orchestration: multi-model routing, context engineering, prompt caching
- MCP: server & client development, multi-host integration
- RAG & retrieval: embeddings, vector retrieval, hybrid fallback
- Eval & observability: guardrails, schema validation, retry recovery
- AI-augmented development workflows
Data & Infrastructure
- PostgreSQL, Drizzle ORM, Docker
- Vercel, Cloudflare, Serverless & Edge functions
Headless CMS & eCommerce
- Sanity, Storyblok, Contentful, Payload, Shopify, Stripe
- Content modeling, multi-environment publishing, migrations
Web & Full-Stack
- TypeScript, Node.js, React, Next.js (App Router, SSR/SSG/ISR)
- TanStack Start/Router, Tailwind, shadcn-style design systems
Experience
PixelPoint Ltd. — Tech Lead & AI Engineer, PrimeUI
Feb 2025 – Present · Remote AI-native website generation platform (primeui.com, launched Feb 2026) building professional-grade, production-ready code. Drove ~70% of platform architecture across AI layer, sitemap canvas, server runtime, public API, MCP server, CLI, code-export pipeline, and component registry.
- Cut client-wireframe turnaround at PixelPoint from a week of Figma prototyping to a few hours: deployed prototypes with real desktop/mobile responsive components, production-level web performance, and dynamic interactions instead of static mockups.
- Designed, built, and published MCP server (
@primeuicom/mcpon NPM, official TypeScript SDK) as the integration layer letting any external codebase (no PrimeUI export required) connect to PrimeUI from inside AI coding environments. 25 tools exposing project management and AI-driven page creation. Tested and tuned across 6 major AI coding agents, including Claude Code, Cursor, and Codex. - Architected end-to-end AI generation layer: multi-model orchestration via Mastra + Vercel AI SDK Gateway (10 agents, 17 workflows), with per-task model routing: Claude Opus 4.6 (Studio chat), GPT-5.2 (outline + section writer), Gemini 2.5 Flash with cached component library (block planning), Claude Haiku 4.5 (short-form tasks).
- Built RAG-based component retrieval: 1,555 markdown chunks across 204 components, indexed with OpenAI
text-embedding-3-small(1536d, 256/32 overlap), cosine top-K with three-tier fallback (exact key → family bucket → semantic). - Built quality guardrails and observability for AI page generation: deterministic rule-based UX scoring for candidate ranking, JSON + props-schema validation with retry recovery (5 retries; triggers: parse, schema, rate-limit), per-step token / latency / cache-hit logging.
- Optimized cost-quality balance: 73% Gemini context-cache hit rate via 24-hour TTL on the component library, ~150K input tokens per page variant across ~9 components and ~5 retrievals, 2–3 min end-to-end latency for the default 3-variant page generation.
- Designed code-export pipeline and CLI: generates a project repo in ~45 sec, ~1.5 min end-to-end from UI Export to a running local project. Output is a production-ready Next.js application with embedded AI tooling, customer-owned and pre-configured for 8 selectable AI coding agents.
- Drove team-wide AI-augmented development: introduced Claude Code from project start, authored team skills and CLAUDE.md conventions, led the team's transition to Codex-based workflow ahead of public release. Built codex-bee (CLI wrapper for long-running Codex sessions) and epic-loop (Codex skill framework with tech-lead and developer agent roles for autonomous large-task-list execution) as a result.
FocusReactive — Tech Lead & Senior Software Engineer
May 2019 – Feb 2025 (5 yrs 10 mos) · London · Remote International web consultancy for clients in the US, UK, Europe, and Australia.
- Architected Tipico US backend platform on Sanity Headless CMS: 5 publishing environments, 3 datasets, white-label architecture for core + state teams. Operating scale: 122 CMS users, 75M requests/month, 98 document types, 165K+ stored documents.
- Led migration of 15+ client projects to Headless CMS platforms (Sanity, Storyblok, Contentful, Payload), restructuring content models and reducing update cycles from weeks to days. Used Claude.ai to analyze inherited codebases, identifying weak points and integration surfaces ahead of migration design.
- Built a Claude-based content automation pipeline: JSON translation flow for 190 country-specific Storyblok pages, integrated into the publishing process and saving ~1 month of manual content work.
- Designed and built CMS-KIT, an open-source Headless CMS starter and component library used across 15+ client projects, reducing kickoff from days to hours and accelerating development by ~30%.
- Led technical hiring and team development: designed an assessment system mirroring real working conditions, conducted 30+ interviews resulting in 8 hires, mentored new engineers; drove team adoption of Claude.ai for research, content, and codebase analysis from 2024 onwards.
GitNation — Senior Software Engineer (Part-time, concurrent)
May 2019 – Feb 2025 · Remote Built publishing flow via Slack API and a reusable web framework for conference websites; migrated GitNation conferences to Headless CMS.
Osome — Frontend Developer
Jun 2018 – May 2019 · Singapore Built fuzzy-logic search system and advanced UI animations for the corporate-services platform.
Skipp — JavaScript Engineer
Nov 2017 – Jun 2018 · Moscow B2B marketplace for Skolkovo Innovation Center (100+ companies integrated); GraphQL-subscription chat for car owners and lessees.
BL Group — Head of Engineering Team
2010 – 2017 · Moscow Led an engineering team designing automation and control systems for dynamic architectural lighting (50+ installations).
Open Source
Storybook — contributor and steering team member during 2016–2017 community revival; addon maintainer (storybook-addon-console — 7M+ NPM downloads; storybook-addon-material-ui — 18K weekly). AI developer tooling — codex-bee, epic-loop.
Education
Master in Engineering, Faculty of Electronic Engineering — Moscow Power Engineering University (MPEI), Russia
Languages
English C1 · Russian native