Jean-Marie Renouard 👋

Principal Data & AI Systems Architect • Creator of MicroLLM • Lead Maintainer of MySQLTuner • Founder of LightPath

I am a Principal Data Architect, Systems Engineer, and Open Source Creator with over 20 years of hands-on experience designing, tuning, and securing mission-critical database infrastructures and high-throughput systems.

Through my engineering consultancy LightPath (founded in 2012), I provide direct, unmediated expertise to international banking institutions, energy giants, national research networks, and large-scale retailers—including BNP Paribas, EDF Datanumia, RENATER, and E.Leclerc GALEC—maintaining strict 99.99%+ availability SLAs across multi-terabyte distributed clusters.

My current engineering focus unites high-performance relational & NoSQL storage engines, cloud-native platform automation (GitOps, Kubernetes, IaC), and deterministic System-1 AI architectures (MicroLLM) delivering certified sub-5ms CPU inference and conformal safety guardrails for autonomous agentic workflows.

The industry-standard, high-performance database diagnostic and configuration tuning engine written in Perl, executed over 15,000 times weekly across millions of production database servers worldwide. As the primary maintainer (#1 contributor worldwide with 1,691+ commits), I have architected its modern capabilities:

• Deep Engine Internals: Comprehensive InnoDB Buffer Pool heuristics, adaptive hash indexes, redo log contention, and thread pool analysis.
• Clustering & High Availability: Galera Cluster quorum validation, split-brain detection, flow control monitoring, and InnoDB Cluster telemetry.
• Modern AI & Agentic Integration: Interactive standalone HTML dashboard reports, machine-actionable JSON output, and native Model Context Protocol (MCP) server integration for automated AI database operations.

# Instant zero-installation live execution
wget http://mysqltuner.pl/ -O mysqltuner.pl && perl mysqltuner.pl

Development Fork: jmrenouard/MySQLTuner-perl

A high-speed, local CPU-only System-1 probabilistic decision and guardrail engine built with NumPy, ONNX Runtime INT8, and SafeTensors. Designed for AI agents and high-throughput API gateways to offload 85% to 92% of deterministic classification and safety decisions in under 5 milliseconds without consuming expensive GPU resources:

• Sub-5ms Inference Pipeline: Combines AnyJev L0 lexical pre-filtering (< 1 ms), calibrated linear heads, and quantized INT8 embeddings on standard x86_64 / ARM64 CPUs.
• Conformal Calibration (RLCD²): Strictly calibrated confidence scores (Expected Calibration Error ECE = 0.008, Brier score = 0.021) with guaranteed coverage $(1 - \alpha)$ and certified abstention (x_abstain).
• Neyman-Pearson Safety Gates: Bounded false-positive rate ($\alpha \le 0.05$) with 97.4% recall against Prompt Injections, PII leakage, and destructive SQL statements (DROP, TRUNCATE).
• Hugging Face Suite: Published suite of 4 official ONNX INT8 models (microllm-fast-cpu-int8, microllm-guardrail-int8, microllm-multilingual-int8, microllm-rlcd2-decision-head) and an interactive Gradio Studio (microllm-decision-studio).

pip install microllm-jev
microllm guardrail "DROP TABLE users;" --policy strict_enterprise

An automated security compliance auditing engine created to enforce Center for Internet Security (CIS) Benchmarks and DISA STIG baselines in regulated, air-gapped environments:

• Pure Python Standard Library: Zero external PyPI dependencies, eliminating supply-chain attack vectors and enabling instant execution on bare-metal and isolated banking nodes.
• Broad DBMS & OS Matrix: 887 automated inspection rules covering MySQL 8.0, PostgreSQL 14/15/16, MariaDB 10.x, MongoDB Enterprise 6.0/7.0, Apache Cassandra 4.x, and Red Hat Enterprise Linux 8/9/10.
• Flexible Execution: Evaluates target hosts via Local shell, Agentless SSH, or containerized Docker inspection, generating actionable executive HTML scorecards and JSON reports.

pip install cis-benchmarks-tools
audit-cis --auto-detect --format html --output report.html

An enterprise-grade Zero-Trust remote task execution and automation platform written in Go:

• Model Context Protocol (MCP): Native MCP server interface enabling AI autonomous agents to securely query system states and trigger maintenance tasks within governed boundaries.
• Zero-Trust Security: Mutual TLS 1.3 encryption, strict role-based access control (RBAC), token authorization, and tamper-evident SHA-256 audit logging.
• Container Presets: Pre-configured, container-isolated maintenance presets for Ubuntu and Red Hat Enterprise Linux infrastructure automation.

• multi-db-docker-env: A modular, containerized multi-engine database sandbox orchestrating concurrent instances of MySQL, PostgreSQL, MongoDB, Redis, and Cassandra with Traefik reverse proxy routing.
• patroni-lab-rhel8: Automated testbed for enterprise PostgreSQL High Availability, leader election, and split-brain mitigation via Patroni, etcd, HAProxy, and PgBouncer on Red Hat UBI 8.
• pg_logical_migrator: Automated Python orchestrator implementing a 17-step sequential lifecycle for near-zero-downtime PostgreSQL migrations via logical replication, available as an official image on Docker Hub.
• easymysql: Operational CLI utility designed for database administrators to provision, clone, and manage multiple isolated MySQL and MariaDB instances concurrently on a single host.

Databases & Storage
AI & Guardrails
Platform & DevOps
Languages & Systems
Cloud & Observability

• Published Technical Book: PHP - Améliorez la performance de vos applications (Éditions ENI, ISBN 978-2-7460-5894-1). An authoritative engineering guide on backend response time tuning, memory profiling, and database access layer optimization.
• MySQLTuner Technical Series: In-depth articles covering MariaDB Galera quorum diagnostics, InnoDB Buffer Pool contention analysis, and AI-ready JSON/MCP outputs on jmrenouard.fr.
• Zero-Downtime PostgreSQL Migrations: Architectural blueprints on parallel physical and logical copy mechanisms utilizing pgcopydb.
• MicroLLM & System-1 Decision Architecture: Research publications exploring conformal prediction calibration, Neyman-Pearson safety theorems, and sub-5ms CPU inference pipelines.
• Conference Presentations: Decks and technical slides hosted on SlideShare.

Outside of solving complex database performance bottlenecks and architecting AI decision engines, my daily energy stems from (in strict hierarchical order):

1. ❤️ My Wife & Children
2. 🤝 My Friends
3. 💻 Coding & Open Source Contributions
4. 🐶 My Dog
5. 🗣️ Endless Technical Speeches & Unbreakable Enthusiasm
6. 🧠 The Million Projects Running in My Head

If my open-source tools or diagnostics saved your production cluster during an outage, helped optimize your queries, or accelerated your pipeline, consider fueling my next late-night coding session!