Serhii Polishchuk

AI-visibility (GEO/AEO) platform · Contract · Remote

• Solo, end-to-end development and full technical ownership of a platform that helps brands get found and cited in AI-assistant answers; in production since November 2025.
• Architecture: TypeScript monorepo, multi-service production deployment, CI/CD.
• Agents: a fleet of 35 specialised agents for visibility measurement, analysis and reporting; two MCP servers (public + admin) exposing 50+ tools to AI clients and internal tooling.
• Backend: Node.js API, background workers and queues, PostgreSQL, Redis, MinIO. Frontend: Next.js dashboard, admin panel, browser extension.
• LLM: 4-provider setup (OpenAI, Anthropic, Google, Perplexity), prompt and pipeline design, and answer-quality evaluation across 100+ monitored prompts.
• DevOps: Railway, Vercel, Docker, monitoring and observability (OpenTelemetry).

AI Founder — startup-validation platform · Project · Remote

• Solo build of a multi-agent platform that validates startup ideas end to end — product, engineering and go-to-market.
• FastAPI backend, Qdrant knowledge base; a cascade of 10+ CrewAI agents orchestrated with LiteLLM and Redis Pub/Sub.
• React frontend with auth and WebSocket live updates; Slack, Reddit, GitHub, GA4 and Stripe integrations.
• Cut research time by ~75% after testing the agent workflows on real users.

Hyperskill · Contract · Remote

• Backend development on the AI-generated-content (AIGC) team: FastAPI and Django.
• Developed and maintained Hyperskill’s open-source testing frameworks (hs-test, hs-test-python, hs-gradle-plugin) in Python and Java — they validate tens of thousands of learner projects.

Hyperskill · Contract · Remote

• Developed the Java/Python testing libraries and authored educational content for the platform.

Freelance (sole proprietor) · Contract · Remote

• Android apps in Java 8 / WebSocket: an SMS-gateway app with backend-driven device management, and a client app.

team-bootstrap

Open-source (v4) role-based AI delivery framework for coding agents: 37 roles, 46 fail-closed checks and 6 pipelines, with an independent verifier, hook-enforced guardrails and OpenTelemetry tracing.

Content-production platform

A 14-stage content-production conveyor on LangGraph with deterministic gates, human approval and a half-price batch lane; driven from a web cabinet, an MCP server or a WordPress plugin; 31 tests.