Housebets
Building the design function from scratch, introducing AI-assisted design and development workflows to rebuild the entire platform and design system.
Brief
Housebets is a crypto casino and sportsbook that had grown quickly without a design function. I joined as the Founding Head of Design to build the function from scratch, establish a design system, rebuild the platform, and reimagine the acquisition and onboarding experiences needed to support its next stage of growth.
Outcome
I led a redesign of the entire front end, starting with a surface-by-surface audit of the site and then building a design system with a shared UI package of 45+ components, shipping across 20 locales to players in 156 countries.
The on-chain bonus campaign became the biggest acquisition launch in the industry, generating 42,031 registrations in its first month, compared with 206–669 in each of the five months before it. Signup completion reached 75.9% in launch week, against a 20–28% baseline.
My role
As Founding Head of Design I own design across the entire experience from the design system, to onboarding and acquisition, and in-house games. I work directly with engineering to take ideas from audit and concept through to production, using AI-assisted workflows to design and build alongside the team.
Beyond the product itself, I built the function around it: establishing the design system, workflows, hiring and onboarding frameworks, and growing the team from scratch.
Toolkit
- Design:
- Figma · Claude Design · Adobe Illustrator
- Build:
- Next.js · TypeScript · shadow-DOM custom elements · packages/ui
- Research & analytics:
- PostHog · Linear · GitHub
01Launching the industry’s largest acquisition campaign
Every casino offers the same generic welcome bonus. We wanted to make ours worth talking about.
The idea was to build a virality-driven acquisition campaign around the players we actually wanted: high-value crypto users with a proven on-chain history. Instead of asking everyone to sign up before showing them an offer, we could read their wallet activity first and give them a bonus based on what they had already done.
That meant the people most likely to be valuable to Housebets could receive the biggest offers, while the campaign itself gave them a reason to share it. The challenge was making that feel instant, trustworthy and completely frictionless.
Turn acquisition into an experience. A personalised offer, built around what players had already done.
I designed the campaign experience and visual assets and the result supported the industry’s largest acquisition campaign. Below is the shipped flow from homepage banner through to the offer credited confirmation.
Banner CTA clicked — 100.0%.
Guest funnel · PostHog
Wallet scan started — 75.9%.
Guest funnel · PostHog
Wallet verified — 37.1%.
Guest funnel · PostHog
Register clicked — 35.9%.
Guest funnel · PostHog
Offer revealed — 34.6%.
Guest funnel · PostHog
Bonus credited — 22.7%.
Guest funnel · PostHog
Guest funnel — first two weeks after launch. The drop from wallet scan to verification is the honest cost of the multi-minute scan and the send-to-prove step. Roughly 30% of everyone who started a scan ended with a credited bonus.
02Turning my hand to game design
We needed a new set of in-house games, but finding the right external partner was proving difficult. So I took on the game design myself, developing the mechanics, interaction models and visual direction from the ground up.
The result is our first set of original House games, with production builds now in development and scheduled to launch by the end of Q3 2026.
Underneath, each game is a shadow-DOM custom element with its own provably-fair engine (HMAC-SHA256), autobet with strategy presets, a stats overlay, hotkeys and a full modal system. They run offline from a single file — a hard requirement so the backend team could swap the money engine in behind a documented adapter seam.
Real production bundles — dice · mines · plinko · keno · blackjack
What can these games do that nobody else’s can?
Bet Burst — exclusive mechanic · hold to chain
Tap Space and you place one bet. Keep it held and, after a deliberate pause, a hint offers a Burst — hold through that and bets chain until you let go. Two things are load-bearing. Bets chain on settle, never on a bare interval, so the engine is never asked to post a bet it is still resolving. And the hold tears down on keyup, blur, tab-hide, modal open and a rejected bet alike, because a keyup swallowed by a focus change would otherwise leave the loop firing with no key held. Manual tab only — on Auto and Advanced, Space already means start-stop autobet. It runs at 320/180/90ms against autobet’s 1000/500/250, because being faster than autobet is the entire point of the gesture.
Lighting — exclusive mechanic · big win
A payout at 10× or more takes the whole card: a radial glow and a floating BIG WIN multiplier, one second, no confetti spam. Small enough to survive a hundred rounds, big enough to feel like something happened.
Drag to paint — exclusive mechanic · Mines grid
On the Mines grid you drag across tiles to select them rather than tapping each one — which meant solving pointer-capture properly, since a naive implementation reports the wrong tile as the drag target and kills the pointer stream mid-gesture.
03Email-first, and pointed somewhere useful
I rebuilt signup around how people actually use the product: making email the primary path, bringing social and wallet options into balance, and sending new players straight into their first useful action: deposit.
- Email and social at equal weight. The previous layout gave Google and wallet the visual priority and pushed email underneath, despite email accounting for 84% of sign-ups.
- Age confirmation where it belongs. 18+ moved out of the terms checkbox into its own explicit consent, with validation that names the missing condition rather than failing generically.
- Then somewhere useful. First-time signups route to the Deposit modal rather than the Wallet page, with methods ordered by real usage and stablecoins pinned. 6,194 players started a deposit in July 2026.
04What the work actually did
From the Housebets PostHog project, the frontend git history and the Linear boards. Each tile names its own source.
Players who completed registration in the on-chain launch month. The five months before ran 206–669.
Acquisition · PostHog · registration_completed · Jul 2026
Signup completion in launch week — 53,433 started, 40,529 finished — against a 20–28% baseline either side.
Flow effectiveness · PostHog · weekly funnel · w/c 20 Jul 2026
Players who started a guest wallet scan in the first two weeks, producing 555,748 scan events.
Reveal reach · PostHog · onchain_bonus_guest_scan_started
Claimed and credited through the flow. A further 379 used the decline path — it needed to exist.
Bonuses settled · PostHog · onchain_bonus_credited
Merged pull requests — 242 commits across 1,726 files. Fourth most active human contributor on the frontend.
Delivery · git · gh pr list --author ismael964
Components in the shared UI package, from 13 pattern and component rebuilds off a full audit.
Design system · packages/ui/src · Linear AFF-214…AFF-332
Players who viewed the XP Race leaderboard in six weeks, generating 49,583 views.
Engagement · PostHog · leaderboard_viewed
Players who saved a rewards slider configuration — the control that previously arrived pre-set and unsaveable.
Rewards adoption · PostHog · rewards_slider_saved
Locales every pattern ships in, including Arabic with full RTL. Players registered from 156 countries.
Reach · apps/web/messages · PostHog geoip
05Weekly signup completion rate
The spike is the on-chain launch. Honest reading: a marketing push drove the traffic, so the defensible design claim is the conversion rate through the flow, not the volume that arrived at it. The launch-week audience also arrived pre-qualified via wallet scan, so it is not perfectly like-for-like with the organic baseline.
06Rebuilding the design system
The product had grown quickly, but the foundations hadn’t kept up. The same patterns were being solved differently across the site, with inconsistent spacing, interaction models and components.
I started by auditing the product as a system rather than a collection of pages. Six pages had six different approaches to spacing. Three sliders used three different interaction models. There was no shared component library, and with multiple agencies contributing to the product, there was no consistent way of working.
The goal wasn’t to document what already existed. It was to work out what should become reusable, then rebuild those patterns properly.
That became the foundation for the House design system: shared tokens, foundations and components that could be used across the product instead of being rebuilt page by page.
The system now covers the core UI patterns across the product, with the same values feeding directly into the builds above. Rebuilt components — 13 shipped: Badge, Link, Footer, RewardsSlider, UserInfoModal, HomepageBanner, CasinoTile, SportsTile, GamePlayingPage, AccountDropdown, SignInFlow, RewardsDropdown, HeaderNav and Tooltip.
Components in packages/ui.
Locales shipped.
RTL support.
Countries registering.
Mobile share of players.
Where my commits went — app/(site).
git
Where my commits went — components/shell.
git
Where my commits went — components/bonus.
git
Where my commits went — public/img.
git
Where my commits went — messages/* (i18n).
git
07The full workstream ledger
Roughly 180 tickets closed across Product Design, Engineering, Tech, Bugs and Arbitrium-Parity.
Design system · Nov 25 — May 26 · Shipped
Full surface audit, then 13 component and pattern rebuilds into a shared UI package of 45 components, live in 20 locales including RTL.
On-chain bonus · Jan — Jul 26 · Shipped
Wallet add and admission errors, queue, scan with archetype scoring, offer reveal, send-to-prove verification including the Polymarket proxy path, qualifying deposit, Share-to-X boost, post-check watcher, credited and four terminal failure screens.
Rewards, VIP & XP · Apr — Aug 26 · Shipped
Rewards slider with a real save state, next-level previews, Benefits section, XP progression fixes, next-claim timing, and the XP Race live leaderboard with campaign integration behind a flag.
Signup & onboarding · Apr — Jul 26 · Shipped
Email-first redesign, equal prominence for email and social, 18+ confirmation, deposit-modal routing, deposit methods sorted by popularity, passkey prompt, blocking iOS and cookie-banner defects.
Mobile & platform quality · Ongoing · Shipped
56% of players are on mobile. Desktop sizing on mobile game layouts, pinch-zoom, bottom sheets, full-page chat, bottom-nav positioning, iOS sheet cropping, plus game loading from white flash to skeleton to branded splash.
Social chat & profiles · Jan — Aug 26 · Shipped
Avatar system, reply-quote behaviour, message bumping, clickable player names across leaderboards and live bets, user info modal with large-number handling, tipping flow redesign.
House games · Jul 26 — now · In flight
Dice set as the reference implementation; Plinko, Mines, Keno and Blackjack brought onto it. Advanced bet ordering, mobile fullscreen, shared bet-info modal, payout hero, Plinko board and rails, Mines drag-to-paint, Keno settle readout, provably-fair modal.
About the games. These are the real production bundles, unmodified — shadow-DOM custom elements with their own HMAC-SHA256 provably-fair engines, autobet with strategy presets, stats overlays, hotkeys and full modal systems. They are running in demo mode against the built-in adapter rather than the live money engine, so balances are play money.
About the numbers. Product analytics coverage only became comprehensive around late June 2026, so feature-level metrics are drawn from a six-to-eight week window and earlier months have gaps. August figures run to the 14th only. The July registration surge was driven by a marketing campaign as well as the flow itself, so the defensible design claim is the conversion rate, not the raw volume.
Not a solo build. The design layer is mine — the audit, the system, the flows, the games, and 205 merged pull requests. Housebets has a full engineering team, who I work alongside with.
External links
Next project: Amazon Autos →