# Nandana Dileep

> AI engineer at Jaguar Land Rover building production agentic systems, LLM infrastructure, and developer tools. Creator of Mycelium, an open-source runtime safety layer for AI agents. IIT Madras '24.

## Identity

- Name: Nandana Dileep
- Role: AI Engineer (GSET), Jaguar Land Rover — Software-Defined Vehicles
- Education: B.Tech, Indian Institute of Technology Madras (IIT Madras), 2020–2024
- Location: Bangalore, India
- Site: https://nandanadileep.com/
- Email: nandanadileep29@gmail.com

## Profiles

- GitHub: https://github.com/nandanadileep
- LinkedIn: https://www.linkedin.com/in/nandana-dileep-06935a212/
- Medium: https://medium.com/@nandanadileep29
- PyPI: https://pypi.org/project/mycelium-runtime/

## Focus areas

- Agentic systems and AI agent runtime safety
- LLM infrastructure and serving (MLX, vLLM-style systems)
- Knowledge graphs and personal AI memory
- AI-native engineering platforms for software-defined vehicles

## Projects

- [Mycelium](https://mycelium-labs.github.io/mycelium/): Framework-independent open-source runtime that prevents predictable AI agent failures before tool execution (LangGraph, CrewAI, custom loops). SDK: mycelium-runtime on PyPI. Source: https://github.com/mycelium-labs/mycelium
- [KAIROS / Auto-Agent](https://github.com/nandanadileep/auto-agent): Autonomous background coding agent with presence-aware notifications and nightly memory consolidation
- [vLLM-style LLM Serving on Apple Silicon](https://github.com/nandanadileep/vllm-like-llm-serving): MLX inference engine with PagedAttention-inspired KV cache, continuous batching, OpenAI-compatible chat completions
- [Identiti](https://identiti.nandanadileep.com): Personal AI memory layer using Neo4j knowledge graphs. Source: https://github.com/nandanadileep/aegis

## Work

- Jaguar Land Rover (2024–present): Core developer on Sarvam, an AI-native Everything-as-Code platform for automotive requirements, HLD/LLD, tests, and compliance; ML systems for test-rig optimization (18% cost reduction on 500K+ daily telemetry events)
- Amazon (Summer 2023): SWE Intern — Alexa latency-analysis service on AWS Lambda, Coral, and OpenSearch

## Writing

- [Checklist to Prevent All the Agent Failure Modes](https://medium.com/@nandanadileep29/checklist-to-prevent-all-the-agent-failure-modes-f4dfdc7597e1)
- [Building a vLLM-Inspired LLM Serving Engine on Apple Silicon with MLX](https://medium.com/@nandanadileep29/building-a-vllm-inspired-llm-serving-engine-on-apple-silicon-with-mlx-65b0576ebd05)
- [Swapped Every Part of nanoGPT With Modern Alternatives](https://medium.com/@nandanadileep29/i-rebuilt-a-tiny-gpt-into-a-mini-deepseek-one-component-at-a-time-718d3e04cfb6)

## Optional

- [Portfolio home](https://nandanadileep.com/): Primary profile page with projects, work, and writing
- [Scraps](https://nandanadileep.com/haphazards.html): Math notes and curiosities
- [sitemap.xml](https://nandanadileep.com/sitemap.xml)