Backend & AI-Systems Engineer · 3 years

I build reliable backends and the AI systems that run on top of them.

I'm Vadim Seliukov, a backend-focused full-stack engineer. For the last three years I've shipped production APIs, data pipelines, third-party integrations, and AI-agent infrastructure. I care about systems that stay correct under real load and real users.

Get in touch Download CV

PythonBackend & APIsData engineeringLLM / agent engineeringAWS · Cloudflare

01

About

I'm a backend engineer who likes the hard, invisible parts: data models that don't rot, APIs that stay predictable, and integrations that keep working after the demo. I work full-stack when a product needs it, but the backend is where I'm strongest.

Most of my last three years has been spent shipping real systems for a live business: booking flows, CRM and calendar integrations, data pipelines, and automations that replaced manual work. A lot of that meant integrating services that ship no official or documented API, analyzing how they behave and building resilient, well-tested clients the business could depend on.

More recently my focus has moved to AI systems and agent engineering, getting language models to take real actions safely, integrating them into existing products, and building the plumbing that makes agents trustworthy. That's the thesis behind Cushio, the reliability-and-safety layer for AI agents I'm building, and it grew out of the same instinct that produced ezRPC, my FastAPI-style framework for HTTP/3.

I studied Computer Engineering at KhNURE and I'm finishing a Master's in the same field (system engineering). I'm most useful on small, senior-minded teams where I can own a system end to end.

02

What I work with

AI AI Engineering & Systems

Building AI agents that take real actions, integrating LLMs into production products, and the reliability plumbing around them.

AI agent engineering LLM integration Tool use / MCP Agent reliability & safety Vector search / embeddings AI automations

API Backend & APIs

Designing services and APIs that stay predictable under load, across sync and async, REST and RPC.

Python FastAPI Django Node.js Express.js NestJS Kafka HTTP/3 · QUIC

DATA Data & Storage

Modeling, pipelines, and syncs across relational, document, and key-value stores.

PostgreSQL MySQL / MariaDB MongoDB DynamoDB SQLite Data engineering Airflow Databricks

SCRAPE Scraping & Integrations

Large-scale data collection and integration with undocumented or unofficial interfaces: protocol & behavior analysis, turned into dependable, well-tested clients.

Web scraping Playwright patchright Undocumented-API integration Resilient clients Integration design

CLOUD Cloud & DevOps

Shipping and running services across clouds, with CI/CD and containers.

AWS Cloudflare Fly.io DigitalOcean Git Docker Kubernetes GitHub Actions

WEB Frontend & Automation

Full-stack when needed, plus workflow automation that connects the tools a business already runs on.

React Next.js TypeScript n8n GoHighLevel

03

Experience

Newly Booked

Aug 2023 - Jul 2026

Full-time · Remote

Full-Stack & Automation Engineer

Owned backend, data and integration systems end-to-end, the breadth of a small senior team in one seat.

Owned features end-to-end across the stack, from the booking-and-operations backend to the interfaces built on top of it.
Built and maintained a suite of 100+ production automations (n8n) across lead intake, call transcription, appointment lifecycle, client onboarding and reporting, removing an estimated 30-50 hours/week of manual operations work (roughly a full-time workload).
Integrated CRM, calendar, booking and telephony platforms (GoHighLevel, Twilio, Monday, Google), including services with no official API, with idempotent syncs and retry/reconciliation logic and resilient clients that survive undocumented changes.
Designed and built a custom, Airflow-like data-orchestration platform from scratch: workers that pull and normalize 13+ sources (GoHighLevel, Twilio, Meta/pixel, Google Analytics, Google Sheets, internal services) into a MongoDB source of truth, behind a load balancer with caching, sustaining thousands of events a day at ~99.8% uptime.
Built the analytics and AI layer on top: dashboards, stats, and vector-embedding search over the data.

04

Selected projects

Open-source project · in development Building

Cushio

A reliability and safety net for AI agents that take real actions. Cushio verifies what an agent actually did (execution receipts), makes repeated actions safe (outcome-aware idempotency), and gives humans a queryable, undoable action ledger. Under the hood, idempotency is keyed on an action's outcome rather than its request hash, the ledger lives in Postgres with compensating-action undo, and it's exposed to agents over MCP.

Python FastAPI PostgreSQL Cloudflare MCP

Open source

ezRPC

A FastAPI-style RPC framework for HTTP/3, built on aioquic with MessagePack, bringing familiar Python ergonomics (routing, typed models, clean handlers) to QUIC. Ships with companion HTTP/3 and HTTP client libraries.

Python aioquic HTTP/3 · QUIC MessagePack Framework design

05

Education

2025 - present

MSc, Computer Engineering

Kharkiv National University of Radio Electronics (KhNURE)

System engineering

2021 - 2025

BSc, Computer Engineering

Kharkiv National University of Radio Electronics (KhNURE)

Bachelor's

Contact

Hiring? Let's talk.

I'm looking for a full-time backend or AI-systems role. If you're hiring, or you just want to compare notes on agents and reliability, send a note. I read every message and reply within a day or two.

contact@seliukov.com linkedin.com/in/vadim-seliukov github.com/kermit-frogs19