Platform Product Manager
Daniel Chao Product Builder
[ 01 — About me ]
I build platforms in some of the most regulated corners of tech, turning a blank page into infrastructure that drives business revenue at scale.
Currently, I'm a product leader at Homebase, where I own Partnerships & Distribution: building the integrations and channel relationships that put our team-management platform in front of small businesses.
[ 02 — Product values ]
How I build
-
01
Build for simplicity
Complexity is a tax every user pays. I cut scope until the core loop is obvious, because a product that needs explaining is a product that needs fixing.
-
02
Ship with quality
Speed matters, but rework is slower than doing it right once. Quality is what earns trust from banks, insurers, and partners who can't afford surprises.
-
03
Own the outcome
Shipping the feature is the start, not the finish. I hold myself to the revenue, adoption, and partner results the roadmap promised.
[ 03 — Craft ]
Top of my craft
B2B API Platforms
I treat APIs like products and developers like users. I've shipped endpoints that hold from the first partner to the hundredth, with real money on every call.
0→1 Product
I find the wedge, then ship the first version myself. I've done it three times, in three industries, talking to real customers every step of the way.
AI-Native
I build with AI, not just around it. I've shipped production agents and working prototypes myself, so I bring engineering something real, not just a spec.
[ 04 — Career ]
I own how Homebase reaches small businesses beyond direct sales: embedded partnerships across the SMB SaaS ecosystem, public APIs, and franchise expansion. The mandate is a 12-to-24-month distribution roadmap and the partner relationships to deliver it.
Zero to one. I built the B2B API platform that turned Gainbridge's direct-to-consumer annuities into insurance infrastructure other companies could embed: a new category, from blank page to revenue.
Built the credit-card processing platform from inception: the APIs and ledger infrastructure that let brands launch credit programs on modern rails.
Ran product on Zelle P2P payments at Chase scale: 22M+ users and $100B+ in annualized volume moving through the products my team shipped.
SOX compliance and internal controls for banks. This is where I learned how regulated institutions actually work from the inside.
Financial advisory and risk consulting. First job, and the foundation for everything regulated that came after.
[ 05 — Case studies ]
Selected work
Chase Contextual Nudges
ML-driven nudges that moved P2P payments back to Zelle.
At JPMorgan Chase, I built the contextual nudge feature in the Chase Mobile app that identified users likely to switch from Venmo and Cash App to Zelle, then served them the right prompt at the right moment. The result: a 6% lift in Zelle monthly active users.
The problem
Millions of Chase customers were running their P2P payments through Venmo and Cash App. Every one of those transactions pulled money out of the bank network, and users paid for it with a 1 to 2 day ACH delay getting funds back into their Chase account. Zelle settled instantly, directly into Chase, and sat right inside the app under Pay and Transfer. Users just weren’t switching on their own.
The insight
The obvious target was users actively sending money out through Venmo: a clear intent signal. But digging into the transaction data with our ML team, the incoming transfers stopped me. A user receiving Venmo payments was waiting up to 48 hours for money to land in their Chase account, paying the cost of someone else’s app choice every single time. The outgoing transfer is an intent signal. The incoming transfer is a frustration signal, and frustration converts faster because the pain is already present. That one insight reshaped the whole strategy: different hook, different message, different emotional path for each group.
How we built it
- I defined the product signals powering the propensity model: transaction frequency, amount, whether the recipient was already on Zelle, time-to-debit after a Venmo cash-out, and repeat sender relationships that indicated a stable payment pod.
- Precision over reach. A loose model nudges people who will never switch, and ignored nudges train users to dismiss everything. We set a 70% confidence threshold: below that, no nudge.
- Staged rollout: two weeks of shadow mode to validate predictions, then a 50/50 holdback pilot. Users who dismissed the nudge and went back to Venmo were deprioritized, so the model learned from rejection instead of nagging.
Why it matters
This is Chase’s “digital everything” strategy made concrete: meeting customers in the moment with data-driven prompts instead of marketing campaigns. Zelle’s trajectory proves the model, since nearly all Zelle activity now happens inside bank apps rather than the standalone app. The lesson: the best conversion target isn’t the user who wants to switch, it’s the user already paying the cost of not switching. Most teams stop at intent signals. The better story was in the data flowing the other direction.
Greenlight Family Cash Card
Credit rails that turn family spending into kids' investments.
As a product manager on Marqeta’s credit platform, I built the card processing infrastructure that powered the launch of the Greenlight Family Cash Card: a Mastercard credit card that pays parents up to 3% cash back and automatically invests those rewards into their kids’ accounts.
The problem
Greenlight had millions of families on its debit card and money app for kids, but nothing for the parents’ own spending. The opportunity was to turn everyday household spend into long-term wealth for kids. That required credit rails, and legacy processors couldn’t support rewards logic this unconventional. Routing cash back into a child’s investment account isn’t a config option on a 40-year-old platform.
My role
At Marqeta, I was the PM building out the credit card processing platform: the issuing, transaction, and program-management infrastructure that lets a partner design a credit product instead of renting a generic one. Working alongside Bend by FNBO’s credit-card-as-a-service stack, that platform is what made Greenlight’s launch possible.
What made the card different
- Up to 3% cash back on every purchase. No rotating categories, because busy parents don’t juggle categories.
- Rewards auto-invest into kid-focused portfolios, converting routine spending into college and savings funds.
- Full integration with the Greenlight app, so credit sits alongside the family’s saving, chores, and financial-education tools.
Why it matters
Covered by CNBC, TechCrunch, CNET, and NerdWallet, the card helped define a new category: family investing credit cards, where rewards route to kids’ futures instead of generic perks. The bigger lesson is that modern issuing infrastructure changes what a credit product can be. When rewards become a design surface instead of a fixed perk, spending becomes a savings engine.
Higher Touch: AI Front Desk for Clinics
Founder-built AI agent for aesthetics clinics.
I’m the founder and product lead of Higher Touch (highertouch.ai), an AI front desk agent for skincare and aesthetics clinics across Southeast Asia. It answers every message, books every appointment, and follows up after every treatment, 24/7, in the clinic’s own language, on the channels customers already use.
The problem
Aesthetics clinics run on real-time messaging, but staff are either with a client in the chair or offline. Roughly 40 to 50% of booking requests arrive after hours, and a customer who doesn’t hear back in minutes messages the next clinic. The owners I’ve talked to aren’t short on demand. They’re short on hours: one told me she spends 3 to 4 hours a day just answering DMs, on top of running the clinic.
The insight
Two assumptions I had to unlearn to design the right product. First, channel: the category got validated when Meta launched its own business messaging agent in 2026, but it’s built for WhatsApp. Thai aesthetics clinics don’t live on WhatsApp, they live on LINE, roughly 56 million monthly active users in Thailand, about 80% of the population, so a WhatsApp-first agent is solving for the wrong market. Second, integration: talking to a clinic operator who’d already deployed an AI chatbot, the bottleneck wasn’t the AI, it was that her vendor tried to replace the clinic management system she already trusted instead of working inside it. Competitors are pitching clinics to migrate onto a new platform. Clinics won’t migrate off software that already runs their business, so the agent has to meet them inside it.
How I’m building it
- Channel-agnostic brain, channel-specific adapters. One AI core owns booking logic, FAQ, and multilingual response. Thin adapters translate it to each channel, LINE first, then web chat, Instagram DM, WhatsApp, and TikTok, so adding a channel is an adapter build, not a rebuild.
- An integrations layer that reads and writes into whatever system the clinic already runs, from a professional clinic management platform down to a plain Google Calendar, so the agent works from real availability and real visit history without asking anyone to switch tools.
- Safety-first by construction: prices, availability, and contraindications are always pulled from a deterministic tool call, never generated from model memory, and a flagged contraindication escalates straight to clinic staff instead of letting the agent decide.
Why it matters
I’m building toward a first Bangkok launch, validating directly with clinic owners before writing production code. In a regulated, trust-heavy vertical, the moat isn’t owning the clinic’s data, it’s the depth of the integration and the judgment layered on top of it. An agent that fits into the tools a clinic already trusts beats one that asks the clinic to start over.