Practical AI engineering by Slava Dubrov
The parts of AI systems that survive production
Agent runtimes, memory, security, retrieval, evaluation, LLM infrastructure — written up the way I’d hand them to the next engineer on the team.
Tracks
Browse all articles →-
AI Agents
Reasoning loops, memory, tools, harnesses
14 articles -
Model Engineering
Fine-tuning, quantization, serving
9 articles -
AI Infrastructure
vLLM, LoRAX, throughput, cost
6 articles -
Retrieval and Language Systems
RAG evaluation, ranking, context
4 articles -
Developer Tooling
uv, pyproject, local LLMs on macOS
8 articles
Engineering the Agentic Stack
In order. Start at the loop.- 01 AI Agent Reasoning Loops: ReAct, ReWOO, Plan-and-Execute 11 min
- 02 AI Agent Memory Architecture: Checkpoints and Vector Stores 33 min
- 03 AI Agent Tool Use: MCP, CLI, Skills, and Code Execution 18 min
- 04 AI Agent Security: Permissions, Sandboxes, and MCP Threats 35 min
- 05 Long-Running AI Agent Runtime: Sessions and Checkpoints 38 min
- 06 Harness Engineering for AI Agents: Designing Control Loops 24 min
Latest
Browse all articles →- TypeScript for Python ML engineers: build an agent service 22 min
Map pydantic to Zod, FastAPI to Hono, and Celery to a Postgres queue, then build a streaming agent service in a pnpm TypeScript monorepo.
- Harness Engineering for AI Agents: Designing Control Loops 24 min
How the control code around an AI agent authorizes tools, preserves progress, verifies completion, and earns each added component through ablation.
- Model Quantization Guide: Foundations to Production Serving 23 min
A practical guide to choosing LLM and diffusion-model quantization paths by bottleneck, hardware, runtime kernel, calibration data, and deploy checks.
One article, every other week
The same posts land on Substack the day they publish. No digest, no drip.