PHASE 1 — THE MOMENT
Render shipped Workflows, a Codex plugin, MCP-triggered deploys, OAuth, CLI lifecycle management, and early-access Sandboxes in five months, with no dedicated designer owning the agent experience layer. Samia Saleem, Head of Design, is hiring the company's first Staff designer for that surface, and she has publicly invited DMs and asked for portfolios.
The real mandate: Build one interaction model that holds across CLI, API, MCP server, SDKs, Blueprints, Skills, documentation, permissions, recovery paths, and the web dashboard. Render's own documentation warns that agents on this surface can modify environment variables and trigger deployments.
Window: Saleem's hiring posts are roughly six weeks old. The role is live on the careers page. She posted about deploying a site with Claude Design within the past 48 hours, so she is publicly active this week and six weeks into a search she is still talking about.
| Dimension | Score | Note |
|---|---|---|
| AI centrality | +2 | Agent infrastructure is a primary strategic bet, not the core product |
| Stage & equity | 3 | Series C extension, $1.5B valuation, pre-IPO, equity in the posting |
| Design influence | 2 | Small team under a Head of Design with org-scaling history; infrastructure ceiling unknown |
| Trajectory | 3 | 100%+ YoY revenue growth, 4× net new MRR since January, 6.5M+ developers |
| Company | 10 | |
| Comp potential | 3 | $218K–$300K plus pre-IPO equity at $1.5B |
| Scope expandability | 3 | First dedicated seat in a category Saleem says is still being defined |
| Craft depth | 3 | Staff IC, hands-on, small team |
| AI exposure | 3 | Agent permissions, trust, supervision, recovery — the core design surface |
| Portfolio value | 3 | Agent experience at a $1.5B infrastructure company; publishable and differentiated |
| Role | 15 |
Narrative fit: The five handoffs in your Trust essay describe a requirement Render has already documented. Agents that trigger deployments and modify state through MCP need human verification, scoped permissions, and a way back. Carrier IQ runs that pattern in production. A published framework plus a working implementation of it is a pairing most candidates for this seat will not have.
Verdict: Act Now. Company 10, Role 15. Both thresholds cleared. The buyer route is the cleanest in the current field: named hiring manager, public invitation to DM, active this week.
PHASE 2 — PORTFOLIO MAPPING
Lead with: Trust essay plus Carrier IQ. The five handoffs — Intent-Setting, In-Progress, Output Review, Decision Gate, Loop Feedback — describe what goes wrong when agents operate on infrastructure. Carrier IQ shows staged work, human authorization, and review in a regulated context. You name the problem before Saleem has to explain it.
Support with: TinyFish as technical currency. You shipped an enterprise web agent platform from zero to production in three months and worked with agent traces, auditability, and governance every day. That keeps the framework from reading as theory. Alibaba is your enterprise-scale anchor ($50B+ GMV, +20% daily transactions) if the conversation turns to platform complexity.
The evidence gap. Nothing published shows you designing across CLI, API, MCP, SDK, and web dashboard as a single product. The posting files that experience under preferred qualifications rather than core requirements, but those surfaces are the job. Treat it as a real filter once she starts opening portfolios, whatever the posting says.
Build or send? Send now. She is actively screening six weeks in, and the trust and supervision angle is genuinely differentiated — developer-tools designers will bring the surfaces but not a published framework for agent permissions and recovery. Your TinyFish API work is defensible as adjacent experience in conversation even though it isn't in the portfolio. If you can build one artifact inside 72 hours — a single interaction model drawn across CLI, API, MCP, and dashboard — it strengthens the pitch. Past 72 hours it costs more than it adds.
Competitive landscape:
- Developer-tools designer (Vercel, Netlify, GitHub pedigree): beats you on CLI, API, and dashboard fluency. Loses on agent trust frameworks and published thinking about agent-operated infrastructure. Confidence: high that this archetype is in the pipeline. Saleem's GitHub tenure means she can evaluate this profile precisely, which also means she knows where it runs out.
- AI-native designer from an AI startup: may have agent interaction work, probably no infrastructure or developer-tools background. Confidence: moderate.
- Enterprise design director: unlikely here. Staff IC seat at an infrastructure company, and Saleem's record suggests she screens for practitioners. Confidence: high this archetype is absent.
- Lead with the framework. The gap is answerable in conversation; do not pre-empt it in writing.
Hardest objection: Has she designed for developers working across programmatic interfaces? The answer is TinyFish, and it is verbal rather than published — an enterprise web agent platform where you shipped API-driven agent tooling into production. Say that plainly, then move to what happens when an agent acts on infrastructure and is wrong, because that is the part your published work covers and the part most substrate-fluent candidates have never had to design.
Buyer context: Saleem's career runs through developer-facing products — Unity, Nike Developer Tools, GitHub — and three concerns recur. Hands-on quality: at GitHub she launched a company-wide weekly production "paper-cut" program. Design-engineering proximity: she helped establish a formal Design Engineer job family there. First-run experience: in a public appearance she argued that an interaction failure during onboarding can undo an otherwise attractive product.
What to ask in conversation: Render's August agent post shows that service-plan changes already require human approval. Ask whether that was a design decision or an engineering default, and whether this hire inherits and validates those trust choices or redesigns them. It signals that you read the product at the permission layer.
Warning sign: Design at infrastructure companies often hits a ceiling where engineering owns the product model and design decorates it. Saleem's org-building history (Nike 3→30, GitHub 60+12) suggests she knows how to fight that. Ask her directly whether design defines the agent permission model or only the interface to it.
PHASE 3 — THE OUTREACH PACKAGE
Confirm the Lab case study pages at junochen.com are resolving before you send anything. As of September 20 the case routes were returning 404 while the standalone apps stayed live. A broken link undercuts the proof if she clicks through.
Warm path: No mutual connections in the record reviewed. This is cold outreach, but she asked for DMs and portfolios in public, so the channel is already open. Don't manufacture a warm path.
First Contact Message
Subject: The trust layer underneath agent deploys
Samia — your post about deploying your partner's therapy site on Render stayed with me. You described setting up a domain, writing redirect rules, watching auto-deploy, and how using the product for something you cared about changed how the experience felt.
That shift is the design problem I've been working on. When an agent runs that same deploy through MCP, the experience turns into a trust architecture question. How does the human know what changed? How do they verify before it commits? How do they recover when it's wrong?
I published a framework for it: five handoffs between human and agent, from intent-setting through output review to loop feedback. I've built three production applications that implement it, including one in a regulated industry where agent-driven automation requires staged human authorization.
I'd welcome 20 minutes to walk through how the framework maps to Render's agent surfaces. You mentioned DMs are open for the Agent Experience seat — happy to continue here or move to email.
— Juno Chen
Resume Emphasis
Open with the Trust essay and Agentic Labs, Carrier IQ first, since it carries the regulated context, the staged authorization, and all five handoffs. Alibaba second for enterprise-scale credibility; surface the +20% daily transactions figure before the GMV number, because the behavioral outcome reads better to a Staff IC panel than the scale number does. TinyFish belongs in the header as most recent role: "Head of Product at TinyFish (enterprise web agent platform, Series A) — shipped the agentic platform from 0 to 1 within 3 months." Thermo Fisher third, attributed as Product Design Director at BCG Digital Ventures, for regulated-environment proof. Cut Red Cross and Equinox+ to single lines; they send the wrong domain signal to this buyer. Don't lead with BCG DV as an institutional brand — lead with the work, carry the title.
Cover Letter Hook
Render shipped agent-operable infrastructure — MCP, OAuth, Sandboxes, CLI lifecycle management — in five months, and the Staff Agent Experience role exists because that surface still has no single interaction model behind it. The five handoffs in "Trust Is the New Interface," implemented end to end in Carrier IQ, are the closest published work I have to the problem underneath this role.
Cover Letter Draft
Render shipped Workflows, MCP-triggered deploys, OAuth, CLI lifecycle management, and Sandboxes in five months. Your documentation notes that the MCP server can perform destructive actions, which means the design surface here is permissions, recovery, and what the human can see while an agent is working. The Staff Agent Experience role exists because that surface needs one interaction model across CLI, API, MCP, SDKs, and dashboard, and it doesn't have one yet.
I've spent the past year building the framework for that problem. "Trust Is the New Interface" maps five handoffs between human and agent — Intent-Setting, In-Progress, Output Review, Decision Gate, Loop Feedback — and I implemented all five in Carrier IQ, a production InsurTech application where agents automate quote workflows under staged human authorization. The pattern transfers directly: an agent that can modify infrastructure state needs a human who can see what it's doing, verify before it commits, and recover when it's wrong. As Head of Design and Research for North America at Alibaba.com ($50B+ GMV, +20% daily transactions), I built trust architecture at platform scale, and the structural problem holds even when the surface changes.
Most recently as Head of Product at TinyFish, I shipped an enterprise web agent platform from zero to production in three months, working daily with agent traces, auditability, and governance in enterprise deployments. I traced agent runs in production to find where the signal broke down between what an agent did and what a user saw.
I moved to product to build AI-natively from zero. I'm returning to design to apply that depth to one specific, high-stakes surface — agent experience on infrastructure. My portfolio at junochen.com holds the Trust essay, three production agent applications, and the enterprise case studies. I'd welcome 20 minutes to walk through how the five handoffs map to Render's agent surfaces.
Post-Outreach Signals
Render behaves like a growth-stage platform, so a reply in 3–5 business days is normal, though her public DM invitation makes faster engagement plausible. If she responds to the framework itself — asks about the handoffs, pushes on one of them — the hook worked; answer inside 24 hours and go deeper. A recruiter-only reply means the message landed but didn't reach her; take the screen and work back to Saleem inside it. If nothing after five business days, send one follow-up hooked to a new product signal from the changelog; they ship weekly. After two unanswered messages, move Render to Watch and spend the time elsewhere.
PHASE 4 — TIMING
| Action | Window | Condition |
|---|---|---|
| Send first contact message | Within 72 hours | Lab case study pages at junochen.com must resolve — verify before sending |
| Produce forward-looking artifact (one interaction model across CLI/API/MCP/dashboard) | 48–72 hours, concurrent with outreach prep | Strengthens the pitch; do not let it delay the send |
| Follow up if no response | Day 5–6 after send | Hook on a new Render changelog entry or product signal |
- Render's Sandboxes early access: The Sandboxes documentation warns that APIs, defaults, and limits may change — if this product ships broadly before your conversation, it becomes a concrete design problem to reference in interview.
- Saleem's adjacent Staff opening: Render's careers page lists a separate Staff Product Designer role covering multi-service architecture, observability, and deployment rollback — understanding how that seat divides from Agent Experience tells you where the boundary of your remit actually sits.
- Render's permission model evolution: The August agent blog post says Render is still developing granular scoped access and guardrails defining what agents may do — monitor whether they ship anything here before your conversation, because it changes whether you're inheriting decisions or making them.
- Dumbleton's AI-prototyping process: Senior Product Designer Brian Dumbleton's portfolio documents a 2025–2026 initiative that turned AI-assisted prototyping into a repeatable process inside Render's production codebase — reading it before the interview gives you a concrete picture of how design currently ships at Render.
