I am a developer focused on making Large Language Models (LLMs), RAG pipelines, and AI Agents faster, cheaper, and more secure. I strongly believe in the Zero-Dependency philosophy: removing bloatware and relying on the pure Python standard library wherever possible.
- 🔭 Currently working on: Expanding my suite of pure-Python AI micro-tools.
- 🌱 Currently exploring: Advanced Prompt Injection defenses and token-efficient data serialization (TOON).
- 👯 Looking to collaborate on: Open-source AI tooling, RAG optimizations, and agentic workflows.
- 💬 Ask me about: Python standard library magic, Prompt Engineering, LLM security, and reducing API costs.
I build ultra-lightweight tools that solve real engineering pain points in AI pipelines.
- 🚑 json-healer-lite: Instantly repairs and parses malformed, truncated, or markdown-wrapped JSON strings from LLMs.
- ⚡ prompt-cache-optimizer: Formats prompt payloads for maximum caching hits on OpenAI and Anthropic to cut API costs.
- 🤖 micro-agent-router: A minimalist, zero-dependency state machine for routing multi-step agent workflows.
- 📝 md-to-text-lite: Strips Markdown syntax and extracts clean, readable plain text for TTS, search indexing, and RAG.
- 📉 json-to-toon-lite: Compresses JSON arrays into TOON format, saving up to 60% LLM tokens.
- 📄 json-to-yaml-lite: AST-based JSON to YAML converter without requiring
PyYAML. - 🤖 function-schema-generator: Auto-generates Anthropic/OpenAI schemas directly from Python functions via
inspect. - 📊 html-table-rescuer: Extracts messy HTML tables and converts them into clean, token-efficient Markdown tables.
- ✂️ prompt-token-minifier: Strips code comments and JSON whitespace to drastically cut LLM API costs.
- 🔗 markdown-link-resolver: Prepares scraped docs & images (auto-base64) for Multimodal LLM ingestion.
"Perfection is achieved, not when there is nothing more to add, but when there is nothing left to take away." - Antoine de Saint-Exupéry