Dmitry Odegov

Senior Software Engineer / AI Engineer

Architecture · Backend · Agentic AI · Full-cycle Engineering

I build production systems at the intersection of backend, architecture, and AI. My foundation is backend development and enterprise-system design; today I specialize in agentic AI: multi-agent systems, RAG, tool calling, AI workflows, and the engineering infrastructure around LLMs. I take a task from business requirements and architecture through implementation, integrations, and frontend to production.

About me in brief

I am an engineer with a strong foundation in backend development and enterprise architecture, while also having built a number of products end-to-end — from requirements and architecture through backend, frontend, integrations, and production.

Today I specialize in production AI: agentic systems, RAG, multi-agent architectures, and AI-assisted engineering, applying the same engineering discipline I developed through backend and full-cycle product development.

How I approach engineering

I start from the task and the system constraints, not from technology.

  1. 01

    Requirements and boundaries

    Formalizing requirements, product boundaries, and constraints: what the system must do ‒ and what it must not.

  2. 02

    Architecture and data

    Service boundaries, APIs, data models, and integrations. Decomposing requirements into maintainable architecture.

  3. 03

    Implementation

    Backend, AI components, frontend, and infrastructure ‒ tools are chosen after architecture, not before it.

  4. 04

    AI as part of the system

    LLMs and agents get interfaces, business rules, tools, constraints, checks, and observability ‒ instead of replacing architecture.

  5. 05

    Production

    Testability, observability, maintainability, and operability matter more than an impressive demo.

This approach lets me use AI to accelerate development without turning production code into a set of unpredictable decisions.

AI engineering, not vibe coding

An engineering approach to AI development

I use AI as part of the engineering process, not as a replacement for engineering work.

In AI development I rely on my backend and architecture background: I define system rules, project structure, component interfaces, business constraints, and quality criteria up front.

For AI-assisted development I use my own engineering harness: a set of rules, architectural patterns, backend/frontend conventions, business constraints, and guardrails that help AI work inside a defined system instead of redesigning it at every step.

Rules before prompts

Business rules, conversation rules, and quality criteria are fixed before models are connected.

A boundary between LLM and deterministic logic

AI generates content and decisions inside an allowed data model; checks, billing, and side effects stay in deterministic backend code.

Observability and evaluation

Call logging, scenario testing, and retrieval/answer quality evaluation are a mandatory part of every AI component.

My engineering profile

One axis: Architecture → Backend & Fullstack → AI Engineering → Production

Architecture

System design, component boundaries, APIs, data, and integrations.

Backend

Python / FastAPI, async services, APIs, business logic, PostgreSQL, queues, caching, and integrations.

Fullstack

Web applications and client interfaces with React/Vue, realtime features, and desktop/mobile delivery using web-based clients and platform integration tools.

AI Engineering

Agentic AI, multi-agent systems, RAG, tool calling, MCP, orchestration, evaluation, and production LLM integration.

Production

Docker, Linux, CI/CD, monitoring, deployment, and operations.

Featured projects

Engineering case studies: product, problem, my role, decisions, and outcome. The stack sits at the bottom of each card.

Featured projects

Multi-Agent AI Support Platform ‒ Beeline

A production-oriented AI support system with multi-agent orchestration, RAG, business rules, and backend API integrations.

A multi-agent AI support system: dialog, corporate knowledge base, and backend tool calls governed by business rules.

Learn more →
Python FastAPI Pydantic-AI RAG Weaviate PostgreSQL Temporal Redis LiteLLM React

Corporate Messenger

A full-cycle corporate messenger built end-to-end: architecture, backend, web client, desktop clients for Windows and Linux, Android client, UI/UX, integrations, and production deployment.

End-to-end product ownership — from architecture and UI/UX to backend, client applications, integrations, and production deployment.

Learn more →
Python FastAPI SQLModel PostgreSQL RabbitMQ Redis WebSocket S3 JWT Vue.js Tauri Capacitor Docker Nginx

AI Landing Page Editor

A full-cycle AI product for generating and visually editing landing pages: architecture, backend, frontend, UI/UX, AI integrations, deployment, and a dedicated Go-based serving layer for published pages.

End-to-end product development, from the editor interface and AI interaction to publishing and serving completed landing pages through a dedicated high-performance rendering and serving layer.

Learn more →
Python FastAPI Vue.js Pydantic-AI LiteLLM OpenRouter PostgreSQL RabbitMQ S3 WebSocket Go Docker Nginx

AI Kitchen Planner

A microservice AI application for kitchen planning: architecture, backend, realtime integrations, and technical leadership of the backend team.

Architectural ownership: microservice boundaries, realtime, queues, AI models, and technical leadership of the backend team.

Learn more → Project presentation →
Python FastAPI PostgreSQL SQLAlchemy RabbitMQ Redis S3 WebSocket Aiogram Docker

Other projects

Aviation Encyclopedia + RAG

Full cycle: knowledge base, ingestion/ETL, embeddings, retrieval pipeline, and RAG over structured sources.

Python FastAPI PostgreSQL pgvector Elasticsearch

Corporate ERP System

Engineering foundation: full business-process lifecycle and hybrid architecture with legacy modernization.

MSSQL C# .NET Python FastAPI

API & Customer Portal

A REST API and customer portal for a wholesale business with data integration and a web interface.

Go React MSSQL Redis JWT

AI Resume Service

A full-cycle AI service: experience analysis, job matching, cover letters, and a web interface.

Python FastAPI PostgreSQL Redis OpenAI API

How my profile was formed

Not a resume copy, but a career narrative: enterprise systems → backend → architecture → full-cycle → AI engineering.

  1. 2018–2023

    Lead Developer ‒ enterprise systems

    C#/.NET → Python → Web → Go → ML

    Legacy modernization, backend, APIs, business processes, architecture.

  2. 2023–2025

    Senior Backend Developer

    Python/FastAPI, microservices, infrastructure

    Microservices, AI/ML products, infrastructure, and technical leadership.

  3. 2025

    AI / Full-cycle Engineering

    RAG, LLM integrations, fullstack

    AI products, RAG, LLM integrations, fullstack delivery from idea to production.

  4. 2026+

    Senior AI Engineer

    Agentic AI, multi-agent systems, RAG, MCP

    Production AI: agentic architectures, knowledge bases, AI engineering harness.

Engineering principles

A senior mindset without “responsible and stress-resistant” clichés.

Architecture before framework

First I define system boundaries and component responsibilities, then I choose tools.

AI is part of the system

LLMs and agents get interfaces, constraints, tools, rules, and observability.

Simple beats complex

I never add a microservice, queue, or AI component just because the technology exists.

Production beats demo

A system must be testable, observable, maintainable, and operable.

Technology serves the task

The stack is chosen for product constraints, not the other way around.

Core stack

The stack confirms the profile instead of being the main content.

Core stack

Python FastAPI PostgreSQL AsyncIO SQLAlchemy Pydantic React/Vue Docker Linux

AI

Agentic AI Multi-Agent Systems RAG Tool Calling MCP LLM APIs Embeddings Vector Search AI Evaluation

Infrastructure

Redis RabbitMQ Temporal S3 CI/CD Monitoring Logging

Additional experience

Go C#/.NET MSSQL MongoDB Elasticsearch Django

Contact Me

Contact Information

Show
tg Show