LAEH@

How I Can Help

Louis-Alexandre Etezad-Heydari·July 28, 2026

I work across many teams on Tenet 2 — agentic products delivering clear economic value. The goal is helping product, eng, and business stay aligned; prototypes are how I express ideas and move them forward. And I always build with Tenet 1 in mind: every product should exercise the harness and generate trajectories for training.

The role: an unsiloed IC reporting to you — serving the vision you've laid out with no turf to defend and no thesis to protect, moving across product, research, and business teams to help them converge on it.

The xWF setup has served this work well — it let me move fast and stay close to the product questions that matter. But xWF status stops at the internal stack: without access to google3 and the tooling around it, a working prototype cannot land where the products actually live, so each one ends as a demo instead of a change. At this point the binding constraint on the work is access, not ideas.

In practice

→Continue with the AI Studio team — evals and system instructions, the UI/UX to operate them, the managed-agents experience — and help align product decisions with the broader goals.

→Apply the same method to every domain you point me at — Antigravity, Workspace, business… Start with the lead's vision; embed with the PMs, designers, and engineers; understand the constraints and what's already on the table; then prototype from the ground up, with the team.

→Build real depth on Tenets 1 and 3 in parallel, from day one — I don't yet understand those layers well enough.

The method — agentic, daily

I don't just use agents — I run an agentic system, daily. My machines share one git-synced harness where agents brief their successors through commit bodies; eight audit agents I wrote gate everything I ship; blind juries across three model families keep my evals from grading their own homework; and my image pipelines run the whole job — multi-model classification through a trained classifier deployed on-device — end to end. The agentic ecosystem your Tenet 1 recipe describes is the world I already operate in.

In evidence

What I bring

Fifteen years on one problem — the interfaces where humans and AI improve each other, since we co-founded Madbits. Prototyping, evals, enterprise deployment, product–eng translation: the same job the whole time.
Enterprise — AI tools shipped across every layer of DDN: single-source pipelines turning scattered material into brand-consistent presentations — web, PPTX, PDF — and sites; role-aware learning paths and messaging drills for sellers and partners; live GPU-cluster data lineage with failure simulation and cost attribution; 4K interactive experiences for NVIDIA GTC — documents, presentations, training, communication, operations: the surfaces Workspace and Gemini Enterprise live on.
Individuals — AI shipped to people with no engineering background: demand forecasting for a five-shop bakery, work-journey optimization for the chimney sweep, admin tools for a village mayor — agents delivering clear value to people who will delete the app the first time it wastes a minute.
Startup — éovision.ai — the AI architecture and UI/UX letting health and life-science professionals train and run their own detection models: draw a few examples, get a working detector, no engineers in the loop — a step toward the automated medicine your doc calls for.
Exploration — the consumer end of the same spectrum: Gemini-powered learning apps for children, generative art and web experiments — the individual end: GApp, AI Studio Build.

Louis-Alexandre Etezad-Heydari·July 28, 2026