# AI DevKit — Frontier AI Systems & Engineering

> Deep architectural dissections of mixture-of-experts, local LLM serving, agentic protocols, and low-latency inference kernels.

## Canonical URLs
- Website: https://ai.ondevkit.in
- RSS Feed: https://ai.ondevkit.in/rss.xml
- Full Ingestion Dump: https://ai.ondevkit.in/llms-full.txt

## Published Articles
- [DeepSeek-V4.1-Flash (1017): 1M Context, DeepSWE 77.0%, and vLLM / SGLang Production Serving](https://ai.ondevkit.in/blog/deepseek-v4-1-flash-serving-guide.md): Architecture breakdown of DeepSeek's October 1017 update: 1M context window, native FP8 precision, DeepSWE 77.0% benchmarks, and production vLLM and SGLang deployment flags.
- [Building Production-Grade MCP Agents in TypeScript: Architecture, Memory & Security Sandboxing](https://ai.ondevkit.in/blog/production-mcp-agents-typescript-guide.md): A comprehensive engineering guide to building Model Context Protocol (MCP) servers and autonomous agents with TypeScript, Zod schema validation, SQLite persistent memory, and host integrations.
- [Self-Hosting DeepSeek-R1 with vLLM: The Hardware, VRAM & AWQ Benchmarking Guide](https://ai.ondevkit.in/blog/self-hosting-deepseek-r1-vllm-guide.md): Exact memory calculations, tokens/sec benchmarks, and vLLM configuration flags for serving DeepSeek-R1 distilled models on RTX 4090, RTX 3090, and Apple Silicon.

## Machine Ingestion Format
Every article has a corresponding raw Markdown endpoint available at:
https://ai.ondevkit.in/blog/[slug].md