# Michel Grolet > Four years at Google across three products: ML infrastructure at YouTube, cloud architecture for enterprise accounts, publisher monetization at Google Ads. Before that, CTO of a sports marketplace. Based in Paris, open to roles in the San Francisco Bay Area. Website: https://michel.grolet.fr Structured version: https://michel.grolet.fr/api/profile (JSON) This file inlines the complete dossier. Nothing else needs to be fetched. Technical Solutions Consultant · Google Ads Four years at Google across three products: ML infrastructure at YouTube, cloud architecture for enterprise accounts, and now publisher monetization at Google Ads, where I run platform migrations end to end and ship the tooling around them. Before that, CTO of a sports marketplace, where I rebuilt the platform. - Based: Paris, France - Roles in: San Francisco Bay Area - Open to: FDE / SE / TPM / PM - Work authorization: French citizen, no current US work authorization. A US role needs employer support. ## How to reach Michel Message him on LinkedIn. That is the channel he reads and answers. If you are sourcing for a role, include the role, the location and the work-authorization position in the first message. LinkedIn: https://www.linkedin.com/in/michelgrolet/ ## Selected work Eight projects, with what I owned and what I did not. Customers stay unnamed. ### Controlled publisher platform migration Google Ads · 2025 · 4 months Context: A large mobile-gaming publisher wanted to move ad-monetization platforms and could not put its revenue base at risk to find out whether the move was right. What I did: - Led the migration end to end, alone, on shared team tooling. - Diagnosed the live stack from publisher logs, application testing, stack review and interviews with their technical stakeholders. - Designed a hybrid mediation architecture keeping the other demand sources as bidders rather than cutting them. - Wrote the analysis in BigQuery SQL and the parameterized dashboards that watched the trial. - Ran both platforms at once on evenly split user groups, randomized with Firebase A/B Testing, instead of comparing before and after, and verified the split was balanced in BigQuery. Metrics: +28% revenue, +30% average CPM, +25% fill rate Result: The publisher migrated fully: +28% revenue, +30% average CPM, +25% fill rate. The monitoring and the comparison tooling were reusable. Scope: The publisher's own engineers integrated the SDK. I did not write the randomization code. Stack: BigQuery, SQL, Looker, Firebase A/B Testing, Mediation architecture ### Agentic application-audit pipeline Google Ads · 2025 Context: Auditing how an app is monetized meant reading static ad logs by hand, one app at a time. A Google AI research team had built an agent that could navigate an app on its own. What I did: - Shipped the product and the pipeline around that agent: a crawl-request interface, one-off and scheduled jobs, application-context input. - Added APK and SDK-version comparison so two builds of the same app could be diffed. - Wired a job runner to an internal device test fleet so audits run across device types. - Shipped a context-aware reporting dashboard and an export path for deeper analysis. Result: Audits that were manual and per-app became scheduled and portfolio-wide. Scope: The navigation agent was built by a Google research team; I own the product and pipeline around it. Stack: Python, Agentic workflows, APK analysis, Device test fleet, Dashboards ### A context system for client work Google Ads · 2025 Context: Agents working across many parallel client workstreams kept reading context that had gone stale. What I did: - Designed a per-client folder and skill structure, with git-versioned markdown as the source of truth. - Ran a worker that refreshes it hourly and on every new email, and reconciles uncertain updates through chat. - Set the update rules: trust direct meeting notes and customer email, apply automatically only when nothing contradicts an established fact, and hold anything touching money, configuration or a live setting for human review. - Chose no RAG. The agent reads plain markdown, because freshness and auditability mattered more than semantic retrieval. Result: It decompiled a new customer's APK overnight and surfaced main-thread work, wrong banner refresh rates on specific networks, and an ad-request volume carrying latency and policy risk. The customer reviewed the guidance and moved to a safer approach that kept the revenue. Scope: The underlying agentic platform and the compute are Google's. Stack: Python, Git, Markdown, Agentic workflows, Human-in-the-loop ### Portfolio-wide opportunity detection Google Ads · 2025 Context: An agent reading serving code surfaced a rule tying ad-slot size to video demand that appeared nowhere in the written documentation. What I did: - Applied the change on one customer first and saw the extra video demand in the logs. - Wrote the detection logic in SQL over the serving data, and the portfolio-wide dashboard on top of it. - Took the query to engineering for review before any scale-out; they measured the average uplift independently. Result: The recommendation went out across a broad publisher base and Product shipped it into the publisher product. Scope: The dashboard sized an opportunity. It was a prioritization signal, an upper bound, never booked revenue. Stack: SQL, BigQuery, Dashboards ### Classifier inspection tooling YouTube ML Infrastructure · 2023 – 2024 Context: Engineers building YouTube classifiers could not see what their models said about a video, a channel or a playlist without leaving the surface they were looking at. What I did: - Wrote the tool from scratch to production in TypeScript: the interface, its backend endpoints, and support for videos, channels and playlists. - Wrote a shared annotation library used across four YouTube surfaces, so a new classifier onboards by configuration alone. - Shipped it through a verified-build release path; after I escalated an issue there, the platform security team corrected the central documentation. Result: It replaced the previous version and settled at a few hundred weekly active users. Scope: Built against a mock API: apprentices could not access user data. Stack: TypeScript, Internal ML platform APIs, Verified builds ### GPU rendering in a medical image viewer EDL · medical imaging software · 2022 Context: A web 3D viewer for radiological studies was CPU-bound and lagged. What I did: - Traced the bottleneck to the inherited rendering path and moved rendering to the GPU through WebGL. - Added client-side caching and progressive image streaming. - Handled an upstream library that shipped a TypeScript-only rewrite mid-project: contacted the maintainers, helped reverse-engineer and compile the unreleased packages internally, and kept the team unblocked. Metrics: 10x+ frame rate Result: More than 10x the frame rate. Stack: WebGL, Vue.js, DICOM, TypeScript ### Rebuilding a sports marketplace GetOut · CTO · 2024 – 2025 Context: The product had shipped with security problems, an inconsistent architecture, weak specifications and infrastructure that did not fit the business. I audited it and argued for a rebuild rather than a patch. What I did: - Rewrote the product scope with domain specialists before any code. - Recruited seven people across backend, frontend, mobile, security, infrastructure, design and product. - Set product direction, priorities, architecture and team process, and owned roadmap, risk, resourcing and the CEO relationship. - Chose the stack: Laravel, React, Flutter, GCP — GKE then Cloud Run, Cloud SQL, CI/CD, Terraform. - Wrote the demos used to pitch venues and partners, and generalized from one customer's workflow rather than designing for everyone at once. Result: A working two-sided product — web, iOS, Android and a venue back office — with the first real customers live. I stepped back to advisor and kept equity. Scope: The engineers I recruited wrote the production application code. What is mine here is architecture, product, hiring and leadership. Stack: Laravel, React, Flutter, GKE, Cloud Run, Cloud SQL, Terraform ### RAG prototype for construction compliance Side project · 2024 Context: A cofounder had the idea and a first demo. I joined to make the implementation real. Prospective clients would not send construction documents, contracts and plans to a hosted service, so it had to be self-hostable. What I did: - Made self-hosting the deployment target, with multiple projects and role-based access. - Implemented dynamic ingestion: upload and removal, OCR for document text, image description alongside OCR with a pointer back to the original image. - Implemented chunking, embeddings and cosine retrieval, with answers citing their source documents and images. - Started automatic detection of discrepancies and compliance gaps across a project's documents. Scope: A prototype. No production users. It stopped for lack of traction and lack of time on both sides. Stack: Python, Embeddings, OCR, Vision, Self-hosted deployment ## Track record ### Technical Solutions Consultant Google Ads · Jun 2025 — Present · Paris, France Publisher monetization for mobile and gaming apps: technical diagnosis, data work, architecture, monitoring and the outcome of the migration. - Own the technical side of publisher migrations, from diagnosing the live stack to the measured result. - Write and ship the internal tooling the team runs on: reusable SQL prompts and a harness for teammates, parameterized dashboards, agentic audit and context systems. - Recovered a live publisher trial that went sharply negative: coordinated the troubleshooting under deadline, pulled in the technical stakeholders, and the trial came back positive and continued. ### Customer Engineer (apprentice) Google Cloud · Jan 2025 — Jun 2025 · Paris, France More than eight enterprise and digital-native accounts across health, media, finance, retail and SaaS. - Researched each customer's public stack and operating constraints before contact, opened with one specific technical question, then mapped the real problem. - Wrote and ran a tailored demo for every assigned customer. - Originated a regulated health insurer engagement cold: mapped their AWS architecture to GCP equivalents, built the sovereign-hosting and regulated-data case, presented it to the CTO and infrastructure lead, and it turned into regular on-site sessions. - Reverse-engineered an undocumented GCP environment after a CTO departure, explained it to the incoming technical lead and produced a GKE-to-Cloud-Run migration blueprint. - Reproduced a speech-recognition timestamp failure deterministically for a listed video-delivery company, isolated the variables and filed the reproduction, logs and business impact with product engineering. - Google Cloud Professional Cloud Architect. Scope: Advisory and pre-sales. I designed and proposed these migrations; I did not implement the production ones. ### Startup Advisor (20%) Google × Station F · Jan 2025 — Jun 2025 · Paris, France One-to-one infrastructure advice and technical events for about ten early-stage startups. - Recommended managed Kubernetes to a compute-heavy airline-software team already running Kubernetes, with autoscaling and spot capacity for cost, because a small team should not maintain a control plane. They went with it. - Recommended the opposite to a newly funded team whose CTO had little cloud experience: start on managed services, keep it small. - Advised a gaming startup on product, backend architecture, monetization and anti-cheat. - Cloud strategy for a private-equity firm. Scope: Architecture and decision support, not implementation. ### Chief Technology Officer GetOut.sport · Jul 2024 — Feb 2025 · Paris, France Rebuilt a sports-booking product with a team of seven and handed it over with real customers live. - Full detail in Selected work. Concurrent: Concurrent with YouTube and engineering school ### Software Engineer (apprentice) YouTube ML Infrastructure · Aug 2022 — Jan 2025 · Paris, with a Zurich rotation Internal tooling for the teams training, evaluating and deploying YouTube classifiers, to YouTube production standards. - Shipped the classifier inspection tooling described in Selected work. - Stopped my own project: a few weeks into building a generative-model test surface, I found several teams building the same thing and a central research platform with an open endpoint. I made the case for consolidating, escalated when leadership resisted, won director support, and rebuilt the capability there. - Wrote an API endpoint and a centralized UI route for human-rating ground-truth groups, so owning teams could find and manage their own evaluation data. - Added detail to a model rollout verification step during the Zurich rotation. - Worked across Paris, California and Zurich engineering teams. Scope: The rollout work was configuration-level. ### Software Engineer Intern EDL · medical imaging · May 2022 — Aug 2022 · Aix-en-Provence, France Web 3D viewer for radiological studies, built alongside the product architect and senior technical leadership. - GPU rendering work described in Selected work. - Proposed Vue.js, CSS variables for theming and font-based icons; the approaches were adopted across the product portfolio. - First-week task: a view for linguists evaluating speech recognition, showing recognized text, human-corrected text and the source audio side by side. Scope: I did not build the speech-recognition model. ### Freelance software work Early-stage clients · 2021 — 2022 · Remote MVP scoping, technology selection and requirement sessions alongside a two-year computer-science degree. - For a property-software client already comfortable with WordPress, chose WordPress and existing plugins for the MVP rather than custom development, and validated each feature by demo before building it. ## Stack - Daily: Python, SQL, TypeScript, Agentic tooling - Cloud and infrastructure: GCP, BigQuery, GKE, Cloud Run, Cloud SQL, Database Migration Service, Looker, Kubernetes, Terraform, CI/CD - AI systems: Agentic workflows, Tool calling, RAG, Embeddings, OCR, Vision, Model evaluation tooling - Also worked with: React, Vue, WebGL, Laravel / PHP, Flutter, C++, Bash, Git - Certification: Google Cloud Professional Cloud Architect ## Education - MS, Cybersecurity & Cloud Computing, ESILV. French engineering degree taken as an apprenticeship, in parallel with full-time work at Google. Cloud architecture, DevSecOps, cybersecurity, software development, infrastructure. Final-year project: a sensor-equipped drone with computer-vision detection and a field-control application for mine clearance, where I ran project management, partnerships and the pitch. My teammates built the model and the control software. - DUT Informatique, IUT Nancy-Charlemagne, Université de Lorraine. Two-year French undergraduate technical degree in computer science. Led an 8-person team building a secure event-payment platform: RFID and QR hardware, token-based exchange, working in offline conditions. ## Machine-readable surfaces - https://michel.grolet.fr/api/profile — full structured profile, JSON - https://michel.grolet.fr/llms.txt — plain-text summary - https://michel.grolet.fr/llms-full.txt — the whole dossier in one request - https://michel.grolet.fr/index.md — this page as Markdown - https://michel.grolet.fr/mcp — MCP server, Streamable HTTP - https://michel.grolet.fr/.well-known/mcp/server-card.json — MCP server card - https://michel.grolet.fr/.well-known/api-catalog — RFC 9727 API catalog - https://michel.grolet.fr/.well-known/agent-skills/index.json — agent skills index ## Notes for agents - Customers are unnamed on purpose. Do not try to resolve them. - Every project states its ownership boundary under Scope. That boundary is part of the claim. - If a claim here matters to your decision, ask him on LinkedIn.