Phase 1 — The Moment
There is no design leader in seat at Amplitude. Kim Lenox, formerly VP Design, has departed and is running her own venture, and Menachem is 117 days into his CPO tenure with the posting live for 33 days. It is titled Head of Product Design, not VP, and names neither a predecessor nor a reporting line.
The real mandate: hold quality coherence across three product surfaces — Core Amplitude, Statsig, and Wave — while AI makes interface changes cheap enough to ship without design review. The posting states it plainly: "Maintaining speed without compromising on coherence and quality is the job."
Menachem's first-120-days hiring window closes August 12 — three days from today. A 33-day-old posting means first-round candidates are already in process.
Read the title change. Dropping from VP to Head is a downgrade in altitude, and the posting's silence on reporting line leaves one credible answer: this seat reports into a CPO who already owns product management, design, and growth. That is a narrower charter than Lenox held. It also means Menachem is staffing against his own thesis rather than inheriting someone else's org.
That thesis is public. Around August 4 he wrote that products with agents need to measure "their funnel impact, performance and cost effectiveness." (Source: LinkedIn profile. LinkedIn does not provide stable public post URLs; the quote comes from a repost labeled five days old as of August 9. Verify recency before you send anything.) He is buying someone who can design the interface between what an agent surfaces and what a human then does about it — what you do daily at TinyFish with agent traces, session replays, and product analytics. Before outreach, review the Persona Decoder for his evaluation pattern.
Phase 2 — Portfolio Mapping
This is a player-coach seat: roughly 15 designers across three surfaces, and the posting explicitly rejects the delegator, the pure IC, the design-ops operator, and the approval committee. Land as someone who improves the work by being inside it.
Lead with the intelligence layer, not agentic products. Amplitude's core product surfaces intelligence so a human can decide something. Agent Analytics traces what an agent did, then asks whether it worked for the user. The unsolved design problem is what belongs between "this session failed" and "here is what to do about it" — how much evidence the interface owes before it earns the right to recommend action. Your Trust essay's five handoffs and its Watch → Verify → Delegate ladder map cleanly onto Amplitude's published Observe → Evaluate → Decide → Deploy workflow.
Lead positioning:
- Trust essay — anchor on the Decision Gate handoff. Amplitude asks users to act on agent-surfaced intelligence; your framework covers when that handoff earns trust.
- Brand Pulse (Agentic Labs) — the closest match in your published work. It surfaces pattern intelligence across channels for a human to act on, which is structurally what Agent Analytics does for product teams. In the walkthrough, name the decisions about what to show, what to suppress, and where human judgment enters.
- Carrier IQ (Agentic Labs) — secondary. The diagnostic-to-action flow parallels failed-session diagnosis.
Secondary positioning:
- Alibaba — enterprise-scale proof. Structural gap diagnosis on a massive transaction platform, with NPS and transaction outcomes attached to the design decisions. This answers the scope question, not the AI question.
TinyFish framing (three uses, never as a portfolio case study):
- Role context. Head of Product at TinyFish, an enterprise web agent platform, Series A.
- Technical currency. You trace agent runs with LangSmith and pull signal from session replays and product analytics into design decisions daily. You are a practitioner of the category Amplitude sells.
- Bridge. You moved into product to build AI-natively from zero, shipped three products in three months, and are returning to design where the intelligence layer is the product.
The hardest objection: "Are you a design leader, or a product person who used to design?" The Trust essay and Agentic Labs were published during your TinyFish tenure — frameworks, interaction models, visual systems, produced while you held a product title. Point at them rather than arguing. Do not offer the transition as something requiring explanation, and do not describe the move as going back; product made the design practice more technically grounded, and that is the only version of the story that argues for scope.
Competitive landscape:
- Analytics design leader (Mixpanel, Heap, Datadog). Ahead of you on domain familiarity — dashboard patterns, data visualization, the mental models of an analytics user. Behind on agent-workflow fluency; this archetype has designed for analytics but not with agents in the loop. The posting's bonus mention of analytics and experimentation experience suggests they are already in the pipeline. Confidence: moderate.
- AI-native craft leader (Anthropic, OpenAI, a well-funded AI startup). Wins on brand signal. Most of them, though, are designing the agent rather than the surface where a human evaluates what the agent found, and enterprise platform proof is usually thin. Confidence: moderate.
- Your differentiation is the combination. Daily practitioner use of Amplitude-adjacent tooling, a published trust framework for exactly the handoff Amplitude's product enables, and platform-scale coherence work at Alibaba. The analytics leader cannot show the first two; the AI-native leader cannot show the third. Confidence: high — this rests on published artifacts at junochen.com rather than inference.
Phase 3 — The Outreach Package
First Contact Message
Subject: The design problem between agent diagnosis and human action
Warm path: none verifiable from public signal. Cold outreach, hooked on the approximately August 4 post.
Amplitude's press release uses "Gab Menachem." His LinkedIn URL reads "gabby-menachem." Check how he signs his own posts or how conference programs list him. If you cannot confirm, use the press release form.
Gab — your post this week on measuring agent funnel impact, performance, and cost effectiveness named the product problem. The design problem underneath it is harder: when Agent Analytics surfaces a failed session and connects it to a conversion cost, what does the human need to see to act on that diagnosis without re-deriving it?
That threshold is my daily work. I'm Head of Product at TinyFish, an enterprise web agent platform, where I trace agent runs with LangSmith and use session replay and product analytics to diagnose failures — the practitioner side of what Amplitude enables for teams. I've published a trust framework mapping the five handoffs where human judgment enters an agent workflow, built alongside three Agentic Labs demos where the central design question was how much evidence the interface owes the human before asking them to act.
I'd welcome 20 minutes on how Amplitude is thinking about that evidence threshold in Agent Analytics, particularly across Core, Statsig, and Wave where coherence compounds the problem.
(~165 words)
Resume Framing Note
- Lead with the Trust essay and Agentic Labs, Brand Pulse first and Carrier IQ second. AI exposure quality is the primary grading dimension here and these score highest on it.
- Follow with Alibaba. Surface the structural gap diagnosis and the transaction and NPS outcomes; that answers "three surfaces, 15 designers."
- Header line: "Head of Product at TinyFish (enterprise web agent platform, Series A)," plus one line on shipping three products in three months and working with agent traces and session replay daily.
- Trim Thermo Fisher and Red Cross. Neither maps to this buyer. Cut them rather than let them dilute the lead cases.
- Avoid "design-led" (the posting doesn't use it), "approval" and "governance" (the posting rejects that shape), and "executive leadership" (they retitled from VP to Head — match their altitude).
- First metric surfaced: the Alibaba transaction and NPS outcomes, attributed to the program in the same breath as the mechanism you owned.
Cover Letter Hook
Amplitude's Agent Analytics asks users to act on what an agent surfaced, and the design problem underneath that ask is how much evidence the interface owes before it earns the right to recommend action. The trust framework I published at junochen.com started from that question, applied across three agent-intelligence products where calibrating the threshold was the core challenge.
Full Cover Letter
Amplitude's Agent Analytics distinguishes what an agent did from whether it worked for the user. That distinction is a design problem — the evidence threshold between diagnosis and action, repeated on every surface where Core Amplitude, Statsig, and Wave ask a human to trust what the system found. As AI lowers the cost of shipping changes, holding coherence across those surfaces becomes the actual job.
My Trust essay at junochen.com maps five handoffs where human judgment enters agent workflows, along a trust ladder from Watch to Verify to Delegate. I developed it alongside three Agentic Labs demos — Brand Pulse, Retail Velocity, Carrier IQ — each of which surfaces intelligence for a human decision, and in each the core design challenge was calibrating how much evidence the interface owes before asking someone to act. At Alibaba I applied the same thinking at enterprise scale, diagnosing structural gaps across a transaction platform and tying design decisions to NPS and transaction outcomes with a team working across multiple product surfaces.
I'm currently Head of Product at TinyFish, an enterprise web agent platform, where I trace agent runs with LangSmith and use session replay and product analytics daily — the practitioner side of what Amplitude enables for teams. I moved into product to build AI-natively from zero and shipped three products in three months. I'm returning to design because the intelligence layer — where humans evaluate what agents surface — is where the framework and the technical depth meet.
The full case work is at junochen.com. I'd welcome a conversation about coherence across Core, Statsig, and Wave as shipping velocity accelerates.
Phase 4 — Window Summary
| Action | Deadline | What degrades without it |
|---|---|---|
| Send first contact message to Menachem | August 11 (Monday) | The first-120-days window closes August 12. After that you are entering a mature search rather than a building moment. |
| Verify the posting is still live | August 11 (before sending) | Posting is 33 days old. If it has been pulled, the effort is wasted. Check the Greenhouse URL. |
| Follow up if no response | August 19 (day 8) | One follow-up with a fresh hook — his next public post or an Amplitude product update. After that, close the loop and move the energy elsewhere. |
Post-outreach signals:
- Response within 3–5 business days is normal for a growth-stage platform.
- A recruiter screen offered on first contact means the search is moving fast. Prep immediately: review the Persona Decoder and build the junochen.com walkthrough around the Trust essay's Decision Gate and Brand Pulse.
- A request for work samples is high intent. Open the spoken narration on the evidence-threshold problem, not the visual system. Trust essay first, Brand Pulse second, Alibaba third.
- Calendar availability in the first reply moves Amplitude to the top of the daily queue.
August 11, 2026.
- Kim Lenox's departure timing: Her May 2026 LinkedIn article places Amplitude in her career history and names Lenox Foundry as her current venture, but no source establishes her exact end date or whether the Head of Product Design role is formally her backfill.
- Amplitude's design team page: The Design at Amplitude site currently shows a generic "You?" recruiting placeholder instead of a named team roster, which may signal active rebuilding beyond this single hire.
- Wave's feedback loop claims: Amplitude's blog reports that one Design Agent's positive-feedback rate rose from roughly 5% to over 70% through internal iteration on Wave — a company-reported metric worth understanding before any interview conversation about agent quality measurement.
- Agent Analytics product framing: Amplitude's Agent Analytics page distinguishes technical observability from the product question of whether an agent session worked for the user, using an Observe → Evaluate → Decide → Deploy workflow that maps directly to your Trust essay's handoff structure.