Влад

AI & FULL-STACK ENGINEER
[відкрити контакти](див. вище в блоці «контактна інформація»)

Vladislav Demianiuk [відкрити контакти](див. вище в блоці «контактна інформація»)

loomarin.com

in/vladislav-demianiuk-a76794295
Works across the whole AI product, from the LLM pipeline to the interface.
Remote
Took an AI SaaS from idea to production as its only engineer.

PROFILE

Engineer focused on LLM systems in production — multi-provider orchestration, RAG, agent workflows and the pipelines that keep them reliable
and within budget. Full-stack in TypeScript (Next.js, Node.js, PostgreSQL), used to owning a feature from data layer to interface, and working AI-
assisted with Claude Code at a production level.

TECH STACK

LANGUAGES TypeScript, JavaScript, SQL, Python

AI / LLM OpenRouter multi-provider orchestration (Claude, GPT-4o, Gemini, Perplexity), open-source models (Hermes), agents & agent
workflows (Mastra), RAG, embeddings, semantic search, prompt engineering, structured output & guardrails, cost control

VOICE AI Deepgram (speech-to-text), ElevenLabs (text-to-speech)

FRONTEND React, Next.js (App Router, Server Actions, SSR), Tailwind, TanStack Query

BACKEND & DATA Node.js, REST APIs, webhooks, Inngest durable jobs, Redis, PostgreSQL, Supabase (RLS, migrations, RPC), Qdrant, payments &
billing

AI-ASSISTED DEV Claude Code, Cursor, Codex — daily production workflow

AUTOMATION Make, Zapier, Inngest, Telegram bots

INFRA & TOOLING Vercel, Railway, Sentry, PostHog, Git

EXPERIENCE

Loomarin Jan 2026 — Present · 7 mos
AI & Full-Stack Engineer Remote · loomarin.com

Sole engineer on an AEO platform measuring how brands get cited inside AI answer engines — ChatGPT, Perplexity, Google AI Overviews.

Multi-provider LLM layer over OpenRouter — model routing, retry/backoff, timeouts and per-request cost accounting, so scans finish even when a
provider degrades.
Retrieval on Qdrant: embeddings, semantic search and entity normalization to match brand mentions across engines.
Scan pipeline on Inngest handling ~90 tracked LLM queries across 3 engines per run, with idempotency and concurrency caps preventing double-
spend.
Structured output that survives model failures — prompt-enforced JSON, a custom parser and a staged validate-and-repair fallback.
PostgreSQL/Supabase data layer (RLS policies, 36 migrations, RPC) and billing: subscriptions, checkout, webhooks.
Dashboard and UX designed and built in Next.js (App Router), React and Tailwind.
Next.js · TypeScript · Node.js · Supabase/PostgreSQL · Qdrant · OpenRouter · Inngest · Vercel · Sentry · PostHog

Raffels Media Sep 2025 — Jan 2026 · 5 mos
Full-Stack & AI Automation Engineer Hybrid

Owned internal tooling and automation for a 30-person team, and led the dev team on internal IT projects.

Custom CRM developed end to end — data model, backend logic and internal dashboards — replacing manual spreadsheet workflows.
Integrations over REST APIs and webhooks connecting the CRM, analytics and internal services into a single data flow.
LLM-powered agents and automations for routine operations: report generation, data processing and task routing.
Telegram bots for internal workflows — notifications, request handling and team commands.
Node.js · TypeScript · PostgreSQL · REST APIs · webhooks · LLM APIs · Make · Zapier

Worksection Nov 2023 — Sep 2025 · 1 yr 11 mos
Technical Project Manager Kyiv, Ukraine · Hybrid

Technical owner of client implementations at a B2B SaaS company, working between clients and the dev team.

API integrations (REST, webhooks) linking the platform to clients' CRMs, messengers, time-trackers and BI tools.
Feature delivery with the dev team: technical requirements, specs, testing and release.
50+ B2B implementations for IT and marketing companies, each with its own technical setup.

PROJECTS

Dataroom dataroom-self.vercel.app

Interactive dataroom app in Next.js and React — UX design, component architecture and motion.

LANGUAGES

Ukrainian Native · Russian Fluent · English Intermediate (B1)