Phase 1: The Moment
Nubank already runs individual AI financial capabilities across its 139M-customer base — credit advice, debt management, spending insights — with individual features reportedly reaching 15M+ monthly active users. It is now building the unified conversational experience that folds those capabilities into a single AI Private Banker product. This hire owns the interaction systems — confidence signals, transparency, error recovery, voice — that decide whether 139 million people act on financial advice from a machine.
The title discrepancy is the first thing to resolve. The Ashby feed labels this "Head of Design, AI Private Banker." The posting body says "Senior Design Manager (M4)." Nubank cut its management levels roughly in half in 2025, and no public source explains where M4 landed in the compressed hierarchy. The posting describes a 6+ person multidisciplinary team, cross-product systems responsibility, and senior stakeholder exposure across the US and Brazil. It says nothing about reporting line, compensation band, budget authority, or who signs off on model behavior and launch criteria.
The search is eight months old. CDO Ethan Eismann promoted this exact role family — manager, lead product designer, content designer — around late 2025 (date inferred from the LinkedIn activity timestamp, not independently confirmed). The August 21 ATS timestamp is a refresh, 8 days old at writing. A management search that runs eight months and resurfaces with the feed label and the posting body disagreeing about level follows the pattern tracked in Issue #8: the committee recalibrated after the first slate failed, and the revised language shows what they learned they needed. Whether the recalibration went up or down — Head confirmed, or Head quietly reduced to a line-manager grade inside AI Core — is what you need to find out before you spend a weekend on this.
The product is partially built. Individual AI capabilities are live and scaled. The unified conversational layer is still moving from definition to launch, confirmed by a current engineering posting (posted August 26, 3 days old). At least one Bay Area product designer has joined AI Core, though the joining date isn't confirmed from the available source. The design leader inherits a live product ecosystem with an active team build underway.
Buyer identification. Eismann is the best-supported contact. He reports directly to founder/CEO David Vélez, promoted these roles personally, and describes leading a multidisciplinary organization that includes PMs, engineers, and data scientists. His essay "Design the System" frames the CDO role as designing five layers, from the design function through operating methods to customer outcomes, with mandatory quality reviews and governance carrying defined veto power. Rohan Ramanath, GM of AI Core, is the model-side counterpart; the posting calls him a "close partner" and avoids reporting language entirely. Ramanath's published research on production AI agents at Nubank's scale treats evaluation quality, human judgment, and online outcomes as one production loop. Either design sets the evaluation criteria inside that loop or engineering hands them down. The posting doesn't say which, and that is the authority question. This is consistent with the earlier analysis in Issue #6, which identified Eismann as the design-side contact.
Rubric — Company (threshold: 10+)
| Dimension | Score | Rationale |
|---|---|---|
| AI centrality | 2 | Primary strategic bet, not core product |
| Stage / equity window | 2 | Post-IPO NYSE, $1B+ quarterly net income, strong growth but limited equity upside |
| Design influence ceiling | 3 | CDO reports to CEO |
| Trajectory | 3 | Accelerating |
| Total | 10 |
Rubric — Role (threshold: 12+)
| Dimension | Score | Rationale |
|---|---|---|
| Total comp potential | 2 | No posted range — scored provisionally |
| Scope expandability | 2 | M4 could be capped — scored provisionally |
| Craft depth | 3 | |
| AI exposure quality | 3 | Trust, supervision, error recovery in financial AI |
| Portfolio value | 3 | |
| Total | 13 | Comp and scope provisional |
Confirm three things: where M4 sits in the post-restructuring hierarchy, who the role reports to (Eismann, AI Core, or neither), and whether design holds authority over evaluation criteria and launch thresholds. If M4 is a senior leadership tier reporting to the CDO, this is a top-five opportunity. If M4 is a line-manager grade inside AI Core product, the role score falls below threshold on comp and scope. Verify before deep tailoring.
Phase 2: Portfolio Mapping
AI Private Banker surfaces financial intelligence — credit recommendations, debt strategies, spending patterns — to customers making consequential decisions about their own money. The design problem sits in the moment between what the AI recommends and what the person does. Ramanath's research shows Nubank already treats evaluation quality and human judgment as one production loop on the engineering side. The Trust essay's five handoffs are the design-side specification for that same loop: which confidence thresholds warrant surfacing a recommendation, which trigger human review, which error states require recovery rather than retry. The posting asks for this vocabulary — confidence signals, transparency, error recovery. Whether the role defines those criteria or implements criteria set elsewhere is what the gate resolves.
Lead with Trust Is the New Interface. Its central claim — users anchor to the worst outcome, not the average accuracy — is the operating principle for financial AI at this customer count. A single confidently wrong debt recommendation costs more trust than a hundred correct ones build, and that is a design failure before it is a model failure. Pair it with Allē (30M members, dual-surface, 3.2× redemption rate) as shipped consumer-scale proof. Redemption rate is the relevant metric here because it measures people acting on what a system recommended, which is the behavioral outcome AI Private Banker has to produce.
TinyFish framing. Most recently as Head of Product at TinyFish, Juno shipped AI-native enterprise tools in production, working through agent traces, auditability, attribution, and governance firsthand. Use it as technical credibility for the trust and evaluation problems this role faces. Past tense. Not portfolio proof.
Objection: enterprise-heavy background against a consumer-first mandate. The screener sees Alibaba, Thermo Fisher, Red Cross and doubts consumer fluency. Allē closes it: 30 million consumer members, dual-surface across provider and patient, $42 CAC, 3.2× redemption — consumer behavior design at the order of magnitude Nubank operates in.
Phase 3: Outreach Package
First contact message — to Ethan Eismann (176 words)
Ethan — I read "Design the System" a while back and the framing has stuck with me: designing the system that designs everything else. It's the problem I see in the AI Private Banker manager role. At 15M+ MAU, the hard question isn't the conversation surface. It's what counts as a trustworthy recommendation — which confidence thresholds warrant surfacing advice, which trigger review, how a wrong answer recovers. Those decisions sit upstream of every model evaluation, and they belong to design.
I've come at this from both sides. I published "Trust Is the New Interface" earlier this year, a framework for human-AI trust built around five handoffs, drawn from shipping agentic products in production as Head of Product at TinyFish. On the consumer side, my closest analogue is Allē: 30M members, dual-surface loyalty platform, 3.2× redemption rate. Designing for people who act on what a system tells them, at that scale.
I'd value 20 minutes on how any of this applies to packaging the existing capabilities into one experience. Worth a conversation?
— Juno Chen
Resume framing note
Open the summary on trust design for AI systems at consumer scale, and surface the Allē metrics before anything else (30M members, 3.2× redemption), followed by the Trust essay as published thought leadership. TinyFish appears as most recent role with production AI depth, past tense: "Most recently as Head of Product at TinyFish, shipped AI-native enterprise tools in production." Push Alibaba and Thermo Fisher down into supporting proof — they establish scale and rigor but shouldn't dominate the first read, and Alibaba's cross-border trust architecture at $50B+ GMV works better as reinforcement on a second pass than as the answer to the consumer question. Drop "enterprise transformation" from the vocabulary. Use "consumer behavior design" and "AI trust systems." If there's a focus section: AI interaction design, trust architecture, consumer-scale product, dual-surface systems.
Cover letter hook (two sentences)
When financial AI reaches fifteen million people, the trust system becomes the product — how confidence is communicated, how errors surface, how someone decides whether to act on what they're told. I've built that system in production at TinyFish and codified the design principles in "Trust Is the New Interface," and I want to apply both to AI-guided financial decisions at Nubank's scale.
Phase 4: Window Summary
| Action | Deadline | What degrades without it |
|---|---|---|
| Send first contact to Eismann; ask about M4 scope and reporting line | September 2, 2026 | Republished search (8 days old) with likely fresh pipeline — first two weeks of a reset are the highest-value window |
| If gate confirms (M4 = senior leadership, CDO or near-CDO reporting): deep-tailor resume and cover letter using Phase 2 mapping | Within 48 hours of confirmation | Tailored application loses advantage as republication pipeline fills |
| If gate does not confirm (M4 = line manager, no model authority): reclassify to Watch | September 12, 2026 |
- Ramanath's production AI paper: Nubank's research on building customer-support AI agents at 100M-user scale documents the evaluation-driven development loop that the design manager role would need to influence — read it before any conversation with Eismann or Ramanath.
- Eismann on AI-first design: His essay "Embracing AI-First Design" argues that AI shifts design value toward problem selection, judgment, and encoding quality into governance — language that maps directly to how he'll evaluate candidates for this seat.
- Adjacent Digital Assets posting: Nubank's Senior Staff Product Designer, Digital Assets role (posted August 27) names Eismann as the portfolio reviewer and GM Michael Rihani as hiring manager, confirming Eismann's active involvement in senior design selection across Nubank's Palo Alto office.
- CPO Rivera's AI-personalization agenda: Carl Rivera's first strategy interview as CPO frames AI-personalized experiences as the company's next expansion vector — useful context for understanding how AI Private Banker fits the broader product roadmap if the conversation advances past the gate.