llm-tool-use

End-to-end retail analytics consulting project built to BCG X standards churn prediction, price elasticity, marketing mix modelling, SHAP explainability, hypothesis testing, and an AI Copilot (agentic tool-use loop) all production Python, no notebooks.

A Model Context Protocol (MCP) server in Go that provides a splunk_search tool, enabling LLMs to query Splunk instances via the REST API

A minimal, retrofittable injection-to-RCE containment loop for tool-using AI agents that accept untrusted file attachments

Tool Governance for Pydantic AI Agents

File-watcher daemon that turns messy markdown meeting notes into clean, well-structured Linear issues. Anthropic forced tool use with a Pydantic-mirrored schema, idempotent batches via sidecar markers, and an eval harness with a no-op chitchat fixture as the hallucination regression catch. Built with the gstack workflow for Kalpi.