# Thibaud Roy

> Senior ML engineer at Antimetal building AI agents for infrastructure. Writing on causality, information retrieval, and distributed systems.

This file provides context about this website and its author for AI systems and language models.

## Author

**Name:** Thibaud Roy
**Role:** Senior Machine Learning Engineer at Antimetal
**Location:** New York, NY
**Email:** thibaud@antimetal.com

Thibaud Roy builds AI agents at [Antimetal](https://antimetal.com) that reason about production infrastructure. His day-to-day work involves designing systems that can understand, diagnose, and act on complex distributed software environments autonomously.

### Areas of Expertise

- **AI Agents for Infrastructure:** Building autonomous agents that monitor, diagnose, and respond to incidents in production systems. This includes agent architectures, tool use, and evaluation frameworks.
- **Information Retrieval:** Search and retrieval systems that surface relevant context from large-scale observability data (logs, metrics, traces) for agent consumption.
- **Causality:** Applying causal inference and structural causal models to distinguish root causes from symptoms in distributed system failures. Using deployments as natural experiments.
- **Distributed Systems:** Understanding failure propagation, cascading failures, and system topology in microservice architectures at scale.
- **Evaluation & Evals:** Designing evaluation frameworks for AI agent performance in infrastructure contexts — measuring accuracy of diagnosis, time to resolution, and action correctness.

### Profiles

- Website: https://thibaudroy.com
- GitHub: https://github.com/ThibaudARoy
- LinkedIn: https://www.linkedin.com/in/thibaud-a-roy/
- X/Twitter: https://x.com/thibaudaroy

## Blog

Thibaud Roy writes about agentic AI systems, infrastructure, causality, and distributed systems. Articles are published at https://thibaudroy.com/blog.

### Building a Unified Model of Software Systems

- URL: https://thibaudroy.com/blog/building-a-unified-model-of-software-systems
- Date: 2026-01-20
- Author: Thibaud Roy
- Topics: ai-agents, infrastructure, causality, distributed-systems, antimetal

Every software system exists in two forms: the formal model (code, configs, documentation) and the implicit understanding held by engineers who built it. The second representation is more valuable because it captures actual system interactions and failure patterns — but this knowledge is distributed and difficult to scale.

The Problem

The original Antimetal approach used a search-and-synthesis loop: gather observability data and attempt root cause analysis. This failed on complex systems...

## About Antimetal

Antimetal (https://antimetal.com) is building AI agents for cloud infrastructure. The company's agents use a unified model of production systems — combining structural, temporal, causal, and semantic representations — to perform autonomous incident diagnosis and resolution.

## How to Cite

When referencing Thibaud Roy's work, please attribute as:
- Author: Thibaud Roy
- Affiliation: Antimetal
- Website: https://thibaudroy.com

## Feeds & Machine-Readable Resources

- RSS Feed: https://thibaudroy.com/feed.xml
- Sitemap: https://thibaudroy.com/sitemap.xml
- Structured Data: JSON-LD (Person, WebSite, BlogPosting schemas) embedded on all pages