Skip to content

AI Agents Engineering: certificate from Prometheus

Українська версія · Certificate timeline page

AI Agents Engineering certificate, Prometheus, 12 November 2026

Today, 12 November, the certificate arrived for AI Agents Engineering on Prometheus. Signed by Dmytro Rashko. The sheet says "attended the course", so this is the course itself, not a separate exam.

Six weeks. Twelve live 90-minute sessions, and a lab after each one. All of it in Go, on Google ADK 2.0.

The program goes like this:

  • Week 1. Models, tokens, and what a request costs. Then an LLM answer as a Go struct with a JSON Schema, instead of text you hope a parser can read.
  • Week 2. A first ADK agent as a graph of nodes. Then one binary and an HTTP service you can put in a container.
  • Week 3. Messy PDFs, chunking, GraphRAG, re-ranking, and a semantic cache. You also see when plain vector search is enough.
  • Week 4. The ReAct loop, and what to do when a tool fails. The error goes back to the model instead of disappearing in the logs.
  • Week 5. Memory on event-sourced state. Several agents, and a pause when a person has to approve the action.
  • Week 6. A coding agent that spawns critics, and self-improvement with a budget cap and one metric. What you keep is an experiment log and a strategy artifact.

Three of these I had already touched separately: agents, RAG, and shipping a small service. Here they sit in one Go graph, which the one-off workshops never did. What the course does not cover: my production traffic, my data, and my incidents. A lab is not an on-call.

Thanks to Dmytro Rashko and Prometheus for the program.