Engineering the Agentic Stack
In order. Start at the loop.
Production AI systems · Berlin
Build AI systems that improve from evidence
I design production AI systems across software engineering, agents, evaluation, retrieval, security, and infrastructure—turning traces and failures into controlled experiments, reliable releases, and safer behavior.
My Recursive Harness Engineering series explores recursive improvement: one agent harness improves another while tests and a person decide what ships.
Automate the experiment, not the authority to declare it safe.
- Staff AI System Engineersoftware · agents · evaluation · security
- 11+ yearsproduction AI and software
- Doctoral degreeAI diagnostics · PhD-equivalent
- Open sourcereproducible systems
Article series
Building and Evaluating Agent Harnesses
Build the harness, then measure each change.
Recursive Harness Engineering
Controlled recursive improvement: one harness improves another.
Other topics
AI Agents 17 Model Engineering 9 AI Infrastructure 6 Retrieval and Language Systems 4 Developer Tooling 8
Most read
- Retrieval and Language Systems NER Guide 2026: GLiNER, spaCy, Transformers, and LLMs 41 min
- Retrieval and Language Systems OCR in 2026: Classical Pipelines, VLMs, and Document AI 22 min
- Retrieval and Language Systems RAG Evaluation Metrics: Retrieval, Reranking, Generation 44 min
- Model Engineering LLM Fine-Tuning Guide: When to Fine-Tune, LoRA and QLoRA 29 min
Latest articles
All articles →- Agent or workflow: five designs for one support assistant 22 min
The same support tasks as an agent, an agent with approval, an agent in a workflow, a router and a workflow with no agent, compared on one test set.
- Designing an agent harness: from task to architecture 28 min
How task requirements shape an agent harness: choosing control flow, tools, code execution, state, permissions and evidence before choosing a framework.
- Recursive harness engineering: from traces to better memory 20 min
Recursive harness engineering in one small loop: an LLM turns a memory tool's wrong answer into a tested fix, and a plain script checks if an LLM was needed.
About
The systems around the model are the product.
I’m Slava Dubrov, a Staff AI System Engineer with a doctoral degree in AI diagnostics. I work across software engineering, agent systems, evaluation, retrieval, security, and production infrastructure.