live demo

This page rewrites itself for whoever is reading it.

The same facts about my work, re-rendered live for a different audience. Pick a perspective — the headline, the description, and the call-to-action regenerate from a language model in real time. It's the engine that runs quietly across the rest of this site, shown here in the open.

select audience
pcherkashin.dev — rendering for: Engineer

AI-First Engineering for Business Transformation

I help companies migrate legacy systems, automate complex workflows, and build AI-powered products using multi-agent orchestration — like an AI engineering team that works 24/7.

generated · run #0
how it works
01

Per-visitor generation

Every block is re-written on demand by a language model for the selected audience — tone, depth, and examples change while the underlying facts stay fixed.

02

Cached before it bills

Each persona/content pair is cached in Upstash Redis, so the model is only ever hit on a cache miss. The cache is the cost gate, not an afterthought.

03

Guardrails on every call

Prompts ship with injection guardrails and the custom-perspective input is sanitised and length-capped — the model can only re-frame the portfolio, never break role.