Splash Sports / Prediction Markets

Building fast prediction-market experiences for sports fans.

Frontend ownership for a sports prediction-market product: UI, state management, animation, product collaboration, and performance work that made market pages feel instant.

Role: Principal Engineer
Stack: React · TypeScript · Zustand · Node.js · NestJS

Splash Sports prediction browsing screen — Browse the markets → compare games → inspect the details.

Splash Sports NFL games screen — Browse the markets → compare games → inspect the details.

01 / Overview

Prediction markets moved from niche to mainstream.

Platforms like Polymarket and Kalshi helped turn prediction markets into one of the loudest product categories in sports, finance, and culture. At Splash Sports, my work focused on bringing that same market-style energy into a fast, understandable web experience for sports fans.

02 / Challenge

Make live markets feel simple, fast, and trustworthy.

The product needed to support real-time market movement without making the interface feel heavy. That meant clear UI, responsive animations, predictable loading, and tight coordination with product and backend engineers while the data model and user flows were still evolving.

03 / Approach

Own the web experience end to end.

I built the web frontend for the prediction-market experience, owning the UI, interaction model, state management, and animation layer. Zustand handled the client state, organized around atomic slices and memoized selectors so each market update touched only the parts of the interface that needed to change.

04 / Solution

Profile first, then make the fast path obvious.

We used profiling audits to find expensive renders, repeated data work, and slow page transitions. From there, I added route prefetching, maintained client-side caching, and shaped the frontend so key market data was fetched once, reused locally, and ready before users needed it.

05 / Impact

A market interface that kept pace with the moment.

The result was a cleaner and much faster prediction-market flow: warmed pages, fewer redundant endpoint calls, smoother state updates, and a web experience that could keep up with live sports behavior instead of feeling like a static contest page.

Performance profile

From market load to instant-feeling interaction.

An illustrative view of the profiling work behind the experience: fewer repeated requests, warmed routes, memoized state slices, and lighter renders across the market flow.

The profiling approach is also available as our open-source Profiling Perf Skill ↗.

10x faster target for profiled page loads

1x endpoint call per cached market view

0 extra re-fetches after local cache hydration

Initial market load Profiling, route prefetching, and cached market payloads reduced the heavy path.

Return to market Client-side cache reuse avoided calling the same endpoint again.

State update cost Atomic Zustand slices and memoized selectors kept UI updates focused.