8 years building cloud platforms and CI/CD pipelines at scale. Today I design and build AI and crypto products in production — agent layer to infrastructure, full-stack. Looking for a role where DevOps, backend, AI and crypto actually meet.
By the numbers
- 0 ● live
merged PRs (since 2025) ● live
across 50+ repos · refreshed hourly via GitHub API
- 0 ● live
PRs reviewed (since 2025) ● live
public reviews on OSS projects
- 0 ● live
stars on contributed repos ● live
sum of top public repos I contribute to
- 20M+
regular viewers reached
Bedrock streaming platform
- 450+
apps on CI/CD platform
Enedis · French national grid
- 1000+
Kubernetes nodes operated
Bedrock streaming infra
GitHub sync · (cached)
Now
Eliza Cloud & white-label AI agents.
I'm a platform engineer on elizaOS Cloud, the multi-tenant SaaS platform that hosts and operates elizaOS agents (open-source runtime, 18k+ ⭐, ~680 contributors). I own the infra there: multi-tenant Postgres isolation with Row-Level Security, a Kubernetes operator that reconciles agents to scale-to-zero, an AWS → GCP cloud migration with no static secrets, MCP servers exposing internal tools, and FinOps governance. I joined as a volunteer in April 2025 and now sit in the top 1% of contributors on the project (212 merged PRs on the core).
In parallel, under an Eliza Foundation contract, I lead US companies deploying their own internal, white-label AI agents with local (on-prem) LLM inference — mostly a technical-steering role.
For what's next, I'm looking for a platform / DevOps / DevEx role where infrastructure, AI workloads and code assistants meet. Available from July 2026.
Updated July 2026
Why hire me
What I bring, concretely.
-
Comfortable with scale
I make decisions for systems that have to stay up under real load — capacity, observability, blast-radius control, performance regression. That's what I did at Bedrock (20M+ regular users, 1000+ Kubernetes nodes) and on the Enedis CI/CD forge serving 450+ applications.
Bedrock · Enedis
-
Full-stack platform mind
I own a vertical end-to-end — the application code AND the infra that ships it. No throwing things over the wall. That's how I've built Go credential rotators, TypeScript backends, Postgres Row-Level Security layers, the Helm charts and Kubernetes operators that ship them, the CRDs, and the CI/CD pipelines around them.
Bedrock · elizaOS
-
AI workloads with SRE rigor
AI workloads deserve the same operational care as any other production system — observability, scale-to-zero economics, regression testing for non-deterministic behavior. That's the lens I brought to Soulmates, where conversational agents run on GKE behind a custom E2E framework: simulated users and an AI judge catching behavioral drift before it reaches users.
Eliza Labs
-
Senior IC who multiplies a team
Autonomous on complex initiatives, and the kind of senior who makes the team around them stronger — mentoring, code review at scale, technical direction. It's how I spent nearly a year contributing open source to elizaOS before being hired full-time, ended up tech lead on Soulmates from day one, ran L3 support for ~10 product teams plus DevOps coaching across 50+ teams at Bedrock, and led EDF Horizon 2030 transformation tracks.
elizaOS · Soulmates · Bedrock · EDF
-
Cost as an architecture signal
FinOps belongs on the platform dashboard alongside latency and error rates, not in a separate finance ticket. I treat cost as a first-class architecture decision. The Bedrock load-testing rebuild was the clearest version of that: ×10 capacity for 90% lower cost — same workflow for the teams, completely different bill.
Bedrock
-
Transparent delivery, defendable estimates
I don't disappear into the code for two weeks and resurface with a black-box result. I keep a live dashboard of where a project stands, communicate often (often more than the team is used to), break work down into estimates I can defend, and balance several streams in parallel without dropping any. Solo or embedded in a team, the rhythm doesn't change.
Across all engagements
Career arc
The career arc, in plain language.
The thread: I build the substrate that lets product teams ship faster — and now, that substrate has to handle AI workloads in production.
-
Lead — internal AI agents (white-label, US) · Eliza Foundation
Remote · freelance
Helping US companies stand up internal AI agents under their own brand, with local (on-prem) LLM inference. A lead / technical-steering role more than hands-on implementation.
- Technical lead and steering for US client companies deploying their own internal, white-label AI agents — direction and pilotage over direct implementation.
- Local (on-prem) LLM inference: agents running on customer-hosted models instead of cloud APIs — data sovereignty and confidentiality as first-class concerns.
- Integration and customization of the agent platform (elizaOS) to each company's internal needs.
- elizaOS
- AI agents
- Local LLM inference (on-prem)
- MCP
- TypeScript
- Kubernetes
- Terraform
-
Platform Engineer — elizaOS Cloud · Eliza Cloud
Remote · Eliza Foundation
elizaOS Cloud — the multi-tenant SaaS platform that hosts and operates elizaOS agents (open-source runtime, 18k+ ⭐, ~680 contributors, 370+ plugins). Joined as an unpaid contributor in April 2025, in the top 1% of contributors (212 merged PRs on the core), hired full-time — now under an Eliza Foundation contract.
- **Top 1% contributor** on elizaOS — 212 merged PRs on the core out of ~680 total contributors. Joined as unpaid OSS contributor in April 2025, hired full-time after a year of consistent shipping.
- Code Assistant products for the team: license and scope management across Codex / Copilot / Claude Code / Cursor / Continue, Continue wired to local LLM inference, and an AI-powered code-review step in CI/CD.
- Internal MCP ecosystem: MCP server for Linear & GitHub issues/tickets, MCP server exposing the Cloud platform backend, an MCP gateway with a dynamic catalog, Composio and an n8n MCP.
- Cloud platform architecture — unified messaging API, multi-tenant Postgres with Row-Level Security, Pepr-based Kubernetes operator that reconciles agents and KEDA ScaledObjects from a custom `Server` CRD — scale-to-zero AI workloads.
- Cloud infra migration from AWS to GCP — GKE Autopilot Terraform modules, Workload Identity Federation for GitHub Actions OIDC, Artifact Registry — no static secrets.
- Gateway tier — direct-to-pod routing via consistent hash ring (Discord), 4 platform adapters (webhook), CNPG Postgres and Redis in-cluster.
- Multi-chain and platform integrations — Solana and EVM (Ethereum, Polygon, Base, Arbitrum, BNB…), Discord, OpenRouter (streaming), WhatsApp. Authored the n8n integration end-to-end (draft/preview/confirm, OAuth + API key bridge).
- Org-wide standardization — logging, TypeScript build pipeline, E2E test infra. FinOps — dynamic per-org rate limits based on cumulative spend, Stripe idempotency, TOCTOU-safe credit deduction.
- TypeScript
- Node.js
- GCP / GKE Autopilot
- Kubernetes
- Terraform
- Helm
- Pepr (operator)
- PostgreSQL / RLS
- CloudNativePG
- Redis
- GitHub Actions OIDC
- Plugin architecture
- MCP
- Composio
- Continue
- Codex
- OpenRouter SDK
-
Technical Lead, Soulmates · Eliza Labs
Remote
Soulmates — an AI-powered dating product, lives entirely on WhatsApp/SMS. No app to install, no profile builder. The whole experience is a conversation with an LLM agent that profiles, matches, and coaches.
- End-to-end technical architecture, from product surface to infrastructure. Tech lead on a 3-person cross-functional team (product + 2 devs), 100+ PRs merged in six weeks.
- Owner of the GKE landing zone (Terraform) — VPC, CloudNativePG, NGINX Ingress, cert-manager, ExternalDNS, SigNoz observability, self-hosted GitHub runners, Artifact Registry — provisioned end to end.
- Application stack — Next.js admin dashboard (funnels, trajectories, matches, safety, conversations), Stripe API for the Connection Fund, WhatsApp Cloud gateway with voice-note STT/TTS via Vertex, containerized async workers (matching + notifications), Postgres schema + Drizzle ORM.
- Designed the full AI pipeline — conversational onboarding, matching, coaching — all LLM-driven.
- Rebuilt the matching engine as a 4-stage semantic pipeline (pgvector top-50 → SQL filters → 6-dim LLM scoring → Gale-Shapley) with per-pair score cache — ~$50/tick → ~$1-2/day.
- Custom E2E framework for conversational AI — LLM-simulated users, AI-judge scoring, full pipeline coverage. Makes non-deterministic agent behavior reliably testable.
- TypeScript
- Next.js
- React
- GKE / Kubernetes
- KEDA
- Terraform
- Helm
- PostgreSQL
- CloudNativePG
- pgvector
- Drizzle ORM
- Redis
- GitHub Actions
- Stripe
- WhatsApp Cloud API
- Vertex AI
-
Full-stack engineer — AI focus · Auto (atnmsxyz)
Remote · freelance
Auto — an AI crypto / DeFi portfolio agent: natural-language intent execution (swaps, cross-chain bridges, Polymarket / Hyperliquid / Aave / Morpho positions) via embedded wallets, on a real-time elizaOS + Convex runtime. A 3-month mission in parallel with Eliza, working in English.
- Agentic architecture rebuild: skill orchestrator and router, action surface, dynamic tool search (SEARCH_TOOLS), preparation / execution / evaluation phases — to make selecting and running dozens of skills reliable.
- E2E test framework with a dynamic judge generated from each skill's features (LLM-simulated user × per skill × LLM judge), fixture resolution via `requiresState`, dual judging gate (execution / quality), per-conversation spend cap and a wallet-write guard in test — a more complete version of the framework I built on Soulmates.
- Multi-chain LLM inference billing: credit ledger, per-loop enforcement gate (atomic hold/release), USDC payment rail (EVM + Solana), auto-refill and dedicated UI — shipped to production.
- Polymarket (V2) work: locally derived deposit wallets, redeem hardening (EOA fallback), cost-basis backfill, burning lost positions, cross-chain bets in a single confirmation.
- TypeScript
- elizaOS
- Convex
- React
- Bun
- Azure OpenAI
- Privy
- Polymarket
- Solana / EVM
-
DevOps Engineer — Core Infrastructure · Bedrock Streaming
Lyon · hybrid · freelance
European streaming platform — 20M+ regular viewers. Embedded in the core DevOps team that runs the entire Kubernetes infrastructure (1000+ nodes) for 50+ product teams shipping continuously.
- Company-wide CLI wrapper shipped to every developer: a single entry point for Terraform, Ansible and GitLab commands, automatic package install, self-update, OpenTelemetry instrumentation and AI summaries of terraform / helm deploy output.
- Full Argo Rollouts integration — progressive canary strategies with automated rollback on business metrics (Apdex, error rate, success rate, custom KPIs).
- Serverless credential rotation system in Go — synced across AWS, Fastly and internal services.
- Hybrid EC2/ECS load testing platform — ×10 capacity at 90% lower cost.
- Go API Gateway for multi-cloud Kubernetes pre-scaling via SQS / EventBridge.
- Infrastructure standardization — Terraform module refactor, GitHub Actions centralization.
- L3 support across 6+ engineering teams. Open-source contributions to ArgoCD and Argo Rollouts (KEDA, Gloo support).
- Go
- AWS ECS / EKS
- Argo Rollouts
- ArgoCD
- Fastly
- Terraform
- GitHub Actions
- Ansible
- OpenTelemetry
-
DevOps Engineer · EDF
Lyon · permanent
EDF — France's largest electricity producer. CI/CD platform serving multiple Java/Angular product teams.
- Built a full CI/CD platform for multiple Java/Angular product teams.
- On-demand environment provisioning via Terraform + AWS Lambda.
- Quality gates with SonarQube, artifact management with Nexus / JFrog Artifactory.
- Observability stack — CloudWatch and Grafana, end-to-end.
- GitLab CI
- Terraform
- AWS Lambda
- SonarQube
- Nexus
- JFrog Artifactory
- Grafana
-
Tech Lead DevOps → Chaos Engineer · Klanik · Enedis mission
Lyon · permanent
Enedis — France's number one electricity distribution network. Critical national infrastructure.
- GitLab CI forge supporting 450 applications — 40+ modular templates, autoscaling runners, Vault, Nexus, automated DRP.
- Industrialized CI/CD across 450+ apps — reusable template library, automated security scans on every pipeline.
- Chaos engineering program — large-scale GameDays (100+ participants) simulating DDoS, database corruption, exposed secrets on resilient EKS clusters.
- Full observability stack: ELK (Elasticsearch / Logstash / Kibana) alongside Grafana and Prometheus, with custom metrics and dashboards; JFrog Artifactory artifact management.
- Scrum facilitation across several teams: backlog/retro planning, sprint planning and estimation sessions.
- Kubernetes / EKS
- GitLab CI
- Vault
- ArgoCD
- Prometheus
- ELK
- JFrog Artifactory
- Grafana
-
DevOps Consultant · Lizeo
Lyon
Lizeo — automotive data company. Multi-project Kubernetes operations.
- Company-wide Ansible playbooks automating the install and configuration of the monitoring tools and the CI/CD chain (Jenkins).
- Multi-project Kubernetes operations across product teams.
- Prometheus / Grafana observability rollout.
- Internal training programs on Docker, GitLab CI and Terraform.
- Kubernetes
- Prometheus
- Grafana
- Docker
- Terraform
- Ansible
- Jenkins
-
DevOps & Software Engineer — four successive roles · Worldline Global
Seclin · apprenticeship → permanent
Three and a half years at Worldline across four successive roles — apprentice, software engineer, then DevOps engineer. Retail and payments products at scale.
- Migrated critical apps to Kubernetes / OpenShift, built Prometheus / Zipkin / Grafana observability stack, automated deployments with Helm + GitLab CI.
- Backend Java / Spring Boot development with full containerization, SonarQube quality gates: doubled test coverage, zero rollback in 18 months.
- Designed and implemented a full CI/CD workflow for a major retail client (apprenticeship, autonomous within a 4-person team), onboarded the team on new tools.
- Mentored apprentices and junior teammates on RUN topics — monitoring, maintenance, pre-prod.
- Java
- Spring Boot
- MongoDB
- Kubernetes / OpenShift
- Helm
- GitLab CI
- SonarQube
- Prometheus
- Grafana
- Zipkin
// SIDE PROJECTS
What I build on the side.
I've been in crypto since 2020 — DCA-ing since the last bear cycle, on-chain well before AI agents made it cool. On the side, I push my own infra and write the code I want to write.
-
Quetalo · AI job-search PWA
Ongoing · personal · live
A solo full-stack, AI-augmented job-search PWA — you swipe missions, bounties and grants like a deck. A deterministic scraper Worker pulls from 120+ French job boards + crypto bounty sources into Cloudflare D1; a Claude routine authors each card and drafts a tailored application per user; semantic matching runs on Cloudflare Vectorize. Three tabs (swipe / pipeline / profile), LLM-based CV extraction, 700+ commits, in production.
React 19 · TypeScript · Tailwind · Cloudflare (Pages/Workers/D1/Vectorize) · Claude routine · OpenRouter · Playwright
-
Flâneur · autonomous AI agent, 24/7
Ongoing · personal
An embodied autonomous AI agent cycling through real cities on a 24/7 livestream, deciding on its own every ~60-90s and launching a city-coin (Solana) per city. "Constrained Freedom" decision architecture (deterministic focus → focus-narrowed LLM schema → post-tick evaluators), an autonomous orchestrator decoupled from elizaOS's message-driven loop, persistent cross-city memory, perception grounded in open sources (OpenStreetMap, Wikipedia, Mapillary). 100+ TypeScript files, 700+ tests.
elizaOS · TypeScript · Gemini (OpenRouter) · PGlite · Geoapify · Mapillary · Solana · Astro 5 / MapLibre
-
CoccoLime · activewear e-commerce
Ongoing · personal · team of 2
A fullstack activewear e-commerce: Astro 6 storefront on Cloudflare (Pages/Workers) + Medusa v2 backend on Railway, PostgreSQL (Neon), Redis (Upstash), R2 storage, Stripe payments (cards, Apple/Google Pay, SEPA, Klarna). Custom Medusa notification module (Resend) and an AI product-image generation pipeline (OpenRouter SDK / vision).
Astro 6 · Medusa v2 · TypeScript · Cloudflare · Railway · PostgreSQL (Neon) · Redis (Upstash) · Stripe · Resend · OpenRouter
-
Sendo · AI crypto advisor
Oct → Nov 2025 · Solana hackathon
Built for a Solana hackathon — an AI portfolio advisor for retail crypto investors. I shipped the core wallet-analysis plugin (multi-DEX decoders for Jupiter, Raydium, Pump.fun, Orca, Meteora, Whirlpool), an async job system with crash recovery, the social-scoring leaderboard with X/Twitter OAuth via Cloudflare Workers, and a Next.js storage-abstraction frontend refactor.
TypeScript · Bun · Solana · Helius / Birdeye · OpenRouter · Cloudflare Workers · Next.js
-
Gotina · multi-chain DCA automation
Ongoing · personal · live (gotina.app)
My own automated dollar-cost-averaging platform. On each recurring USDC deposit, an orchestrator runs a 5-phase swap/bridge/deposit pipeline across Solana, Bitcoin, Ethereum and Base (Jupiter, Kamino, Chainflip, Mayan, Wormhole CCTP, Morpho), routing assets to cold-storage Ledger addresses. Per-execution Durable Object state machine with idempotent recovery, a gas-aware fee floor and auto SOL refill, SIWS auth and Telegram notifications. Multi-tenant v2 on Cloudflare Workers.
TypeScript · Bun · Cloudflare Workers & Durable Objects · Hono · Wormhole CCTP · Jupiter · Kamino · Chainflip · Morpho · Turso · Next.js / Astro
Case studies
Deep-dives on real work.
Context, decisions made, what I’d do differently. Honest, structured, 5–8 minutes each.
-
Bedrock Streaming · 8 min
Progressive delivery at Bedrock — Argo Rollouts for 50+ product teams
Rolled out Argo Rollouts as the standard progressive-delivery framework for 50+ product teams. Replaced health-check gates with metric-based gates (Apdex, error rate, custom KPIs). Contributed upstream support for KEDA + Gloo Gateway.
Read
-
Bedrock Streaming · 7 min
Bedrock load-testing platform — 10× capacity at 90% lower cost
Rebuilt Bedrock's load-testing platform on a hybrid Amazon EC2 + ECS architecture integrated with Gatling Enterprise Cloud. Result: ~10× more capacity at ~90% lower cost — by treating FinOps as an architecture problem, not a finance ticket.
Read
-
Enedis · via Klanik · 8 min
DevOps at scale on critical infrastructure — GitLab CI forge for 450+ apps & Chaos Engineering GameDays at Enedis (via Klanik)
At Enedis, France's primary electricity distribution operator, I helped build from scratch a Kubernetes-based GitLab CI forge serving 450+ applications, then co-designed large-scale Chaos Engineering GameDays simulating DDoS, database corruption and secret leaks. The common goal: make teams anticipate failure instead of reacting to it.
Read
Stack & architecture
The stack I run, end to end.
A request hits the top of this stack and travels down. Hover any tile to see exactly where I put it to work.
AI / agents
LLM-driven products in production
- elizaOS
- Anthropic API
- OpenAI / OpenRouter
- Agent design
- RAG / memory
Backend & data
Where the application logic lives
- Node.js / TypeScript
- Go
- PostgreSQL · RLS
- Redis
- Drizzle ORM
Platform & delivery
How code reaches production safely
- Argo Rollouts
- ArgoCD
- GitHub Actions
- GitLab CI
- Helm / Charts
Infrastructure
The substrate everything runs on
- AWS · EKS
- GCP · GKE
- Kubernetes
- KEDA
- Terraform
- Cloudflare
Projects & organizations
Where my commits land. ● live
These numbers are fetched live from the GitHub API on every visit, cached for one hour.
-
elizaOS/eliza
19kOpen source agentic operating system
TypeScript 775 commits 323 PRs -
elizaOS/eliza-archive
0Archive of legacy elizaOS branches, tags, and releases (v0.x and v1.x). The main repo is at https://github.com/elizaOS/eliza
TypeScript 375 commits -
standujar/plugin-composio
9A powerful ElizaOS plugin that integrates 250+ external tool integrations
TypeScript 58 commits 1 PRs -
elizaOS/elizaos.github.io
107Leaderboard of Eliza Contributors
TypeScript 35 commits 9 PRs -
Sendo-labs/plugin-sendo-worker
1TypeScript 26 commits 6 PRs -
Sendo-labs/plugin-sendo-analyser
2TypeScript 23 commits 13 PRs
Organizations
+ 2,545 additional contributions on private repositories.
Featured open source
A handful of merged PRs.
- #6167 feat: Entity RLS Infrastructure + API Clarity + PerformanceelizaOS/eliza
- #6101 feat: Add PostgreSQL Row-Level Security (RLS) multi-tenant isolationelizaOS/eliza
- #6095 feat: implement unified messaging API with elizaOS.sendMessage()elizaOS/eliza
- #5864 refactor: Multi-Agent Architecture Refactor with ElizaOS Core and CLI cleanupelizaOS/eliza
- #6217 fix: encryption for character secrets in correct orderelizaOS/eliza
- #6169 refactor: Standardize Logging Across Core, CLI, and ServerelizaOS/eliza
- #6201 feat: Unified API - serverless - nodejselizaOS/eliza
- #6060 feat(cli): Simplify CLI to use server / coreelizaOS/eliza
Skills
-
Languages
- TypeScript
- Go
- Python
- Shell
- Java
-
Cloud & infra
- AWS
- GCP / GKE
- Kubernetes
- KEDA
- Terraform
- Helm
- Docker
- Argo Rollouts
- Ansible
-
CI/CD & DX
- GitHub Actions
- GitLab CI
- ArgoCD
- Dagger
- Bun
- Vitest / E2E
- Jenkins
- JFrog Artifactory
- Nexus
- devcontainers
-
Backend & data
- Node.js
- PostgreSQL
- Redis
- Drizzle ORM
- REST / API design
- Convex
-
AI / LLM
- elizaOS
- Anthropic API
- OpenAI API
- Agent design
- RAG / memory
- Codex
- Continue
- MCP
- Composio
- n8n
-
Observability
- ELK
- Prometheus
- Grafana
- OpenTelemetry
- SigNoz
- New Relic
-
Web3 / crypto
- Solana
- EVM
- Privy
- Viem
- Jupiter
- Polymarket
-
Frontend
- React / Next.js
- TailwindCSS
- Astro
About
Who am I?
Available from July 2026
Open to platform · staff DevOps · AI infra roles
Eight years of platform and DevOps work, two of them embedded with AI engineering teams. I sit at the intersection where infrastructure, application code and AI workloads have to ship together — and where most companies still don't have a clear playbook.
What I'm looking for next
Book a 30-minute intro call-
The role
Autonomous, with a tech-lead reflex
I want to own a topic end-to-end: set the technical direction, mentor the juniors around me, translate complex tech into something the product team can act on, and run the rituals — planning, Scrum, realistic deadlines.
-
The ground
Topics that need to be built, scale up, and matter
A construction phase, on platform or backend — including projects that lean heavier on application code than on DevOps. Topics that have to scale, that matter to the product, and where optimization — cost, performance, infrastructure — is a first-class concern.
-
The setup
Remote anywhere, or hybrid in Montpellier
Fully remote (Europe, US, anywhere with sane time-zone overlap), or hybrid based in Montpellier. Occasional travel within France works fine. A team with a flat hierarchy that lets engineers do their work.
Contact
Let’s talk.
Best path: a 30-minute call. A detailed email works just as well.