Sergey Lyapustin | AI Architect

Let's talk AI

AI Architect with 15+ years in software engineering. Language-agnostic, goal-focused. I design intelligent systems, lead engineering teams, and bridge the gap between AI research and production. From conversation intelligence to agent architectures — I turn complex AI into real-world products.

@slyapustin

Get in Touch View Projects

About

Engineering with purpose

I'm an AI Architect. Language-agnostic and goal-focused — I pick the right tool for the job, not the other way around. Over the years I've designed and shipped AI-powered platforms, LLM integrations, and production ML pipelines for startups and enterprises alike.

I care about outcomes over technology choices, clean system design, and infrastructure that scales. When I'm not building AI systems, I maintain open source projects and explore the frontiers of generative AI, agent architectures, and cloud-native ML tooling.

Open Source Projects

360+

GitHub Stars

Primary Focus

Python

Core Language

Tech Stack

Tools of the trade

Technologies I work with daily and bring to every project.

AI / ML

LLMs RAG AI Agents LangChain PyTorch OpenAI Claude API

Backend

Python Django FastAPI Celery Airflow WebSockets

Infrastructure

Docker Kubernetes AWS PostgreSQL Redis Git

What I Do

Focus areas

Where I bring the most value.

AI Agent Architecture

Designing and building production-grade AI agents — from conversation flows and tool orchestration to reliability, cost optimization, and human-in-the-loop patterns.

Conversation Intelligence

Building systems that understand, analyze, and act on conversations at scale. Real-time transcription, intent detection, sentiment analysis, and actionable insights from voice and text.

LLM Integration

Integrating large language models into existing products and workflows. RAG pipelines, prompt engineering, fine-tuning strategies, and MCP server development for tool connectivity.

Now

Currently exploring

MCP Servers vCon Standards AI Agent Frameworks Conversation Intelligence Supabase + pgvector Claude API & Tool Use Enterprise AI Adoption

Open Source

Projects

Selected repositories — from AI tooling to battle-tested Django libraries.

datadog-mcp

MCP Server for Datadog — comprehensive monitoring and observability tools via Model Context Protocol.

MCP-Salesforce

MCP connector for Salesforce — enabling AI agents to interact with CRM data through the Model Context Protocol.

fixly

AI-powered text fixer Chrome extension — grammar, tone, and style corrections using LLMs.

django-classified

Full-featured classified ads application built with Django. One of the most starred Django classifieds projects.

django-chatbot

Real-time chatbot with Celery, Django Channels (WebSocket) — the foundation for conversational systems.

django-s3-like-storage

Your own Amazon S3-compatible Django storage backend for self-hosted object stores.

Blog

Thoughts & writing

Notes on AI, engineering, and building things that matter. RSS

Apr 2026 · 10 min read

Your Second Brain Has Amnesia

Three ways to give AI persistent memory — Obsidian plugins, Claude Code's AutoMemory, and Karpathy's LLM-compiled wiki. Plus 16 practical tips.

Obsidian LLMs PKM Claude Code Memory

CLAUDE_CODE_NO_FLICKER: The Fix a Year in the Making

Terminal flickering was Claude Code's most upvoted issue for over a year. One environment variable finally fixes it.

Claude Code Developer Tools Terminal

When Building Is the Easy Part

AI collapsed the cost of building software. The value is shifting from implementation to strategy, domain expertise, and distribution.

Strategy AI Engineering Leadership Product

Supabase: The Database AI Agents Deserve

Why Supabase is quietly becoming the best database platform for AI agent applications — real-time subscriptions, pgvector, RLS, and Edge Functions ...

Supabase AI Agents Database Architecture

Building AI Agents That Actually Work

Lessons from shipping production AI agents — reliability, cost, trust, and the architecture that survives contact with real users.

AI Agents LLMs Production

RAG Is Not Enough

RAG got us far, but some problems need more. Exploring fine-tuning, tool use, and hybrid approaches for complex AI systems.

RAG Architecture

The Language-Agnostic Engineer

Why I stopped identifying with a single programming language and started focusing on outcomes. A case for tools over tribes.

Career Engineering

How I Built My Own Amazon S3 Storage

A deep dive into building a self-hosted S3-compatible storage solution with Django and MinIO.

Django S3 Infrastructure

Read on Medium Jul 2019

Heroku Chatbot with Celery, WebSockets, and Redis

Building a real-time chatbot using Django Channels, Celery for async tasks, and Redis as the message broker.

Django WebSockets Redis

Read on Medium Nov 2018

Django Polls App with Elasticsearch

Adding powerful full-text search to a Django application using Elasticsearch and django-elasticsearch-dsl.

Django Elasticsearch

Read on Medium

Contact

Let's work together

Have a project in mind or looking for a consultant? I'd love to hear from you.

GitHub Email LinkedIn Medium