Skip to main content
Tinkerer is an automated experimentation system that uses LLM-guided tree search to solve Kaggle competitions and algorithmic optimization problems.

How It Works

Tinkerer Framework Architecture
The system iteratively generates solutions, implements them via coding agents, evaluates results, and uses feedback to guide the next iteration.

Supported Benchmarks

Generic Problems

Solve arbitrary problems with custom evaluators and stop conditions.

Core Components

ComponentDescription
OrchestratorAgentCoordinates the solve loop, manages budget (time, cost, iterations)
Search StrategyTree search for solution exploration (expand, select, prune)
Experiment WorkspaceGit-based workspace with branch per experiment
Coding AgentsPluggable code generators (Aider, Gemini, Claude Code, OpenHands)
Knowledge SearchNeo4j-based domain knowledge retrieval