Dmitrii Belikov AI and finance

English Русский

Courses

Modules · 8

Claude Code, end to end

  • Claude Code
  • Agents
  • Tooling

Ask about the next cohort

Who it is for

Analysts, quants, and engineers who already write code and want an agent to carry the parts that are mechanical without losing control of the parts that are not. No prior agent experience assumed; comfort with git and the command line is.

How it runs

Each module is a short read plus an exercise you run against your own repository. Nothing is demonstrated on a toy project — the point is to leave with the tool configured for the work you actually do.

  1. Getting set up

    Install · auth · first session

    Installing the CLI, authenticating, and choosing what the agent is allowed to do before it does anything. Ends with a working CLAUDE.md for one of your own repositories, written by the agent and corrected by you.

  2. How the agent reads a codebase

    Search · context · orientation

    Why pointing at the right file beats describing it, when to let the agent explore and when exploration is waste. The mental model here decides whether everything after it feels fast or frustrating.

  3. Editing with a safety net

    Plan mode · diffs · git

    Reviewing a plan before any file changes, reading diffs quickly, and keeping git history clean enough that reverting is cheap. Practise on a branch you are willing to throw away.

  4. CLAUDE.md as an instrument

    Conventions · house style

    What belongs in project memory and what quietly makes output worse. Encoding the conventions a new colleague would need, and nothing else.

  5. Slash commands and skills

    Packaging repeatable work

    Turning a workflow you repeat into something you invoke by name. Where a slash command is enough and where a skill with its own instructions pays off.

  6. Hooks and permissions

    Guardrails in settings.json

    Making the harness enforce what you would otherwise have to remember: formatting on save, tests before commit, commands that never run without asking.

  7. Connecting your own data

    MCP · external tools

    Wiring the agent to the systems your work actually lives in — databases, warehouses, issue trackers — and deciding what it should never be able to reach.

  8. Finance work specifically

    Notebooks · data · secrets

    The habits that matter when the code touches money: reproducible pipelines, notebooks that survive review, and a discipline for credentials that does not depend on anyone being careful.

Ask about the next cohort

← Back

01/01 · Courses0%Modules · 8