Building a Scalable, Data-Driven Biome System
World scaleEcology awareArtist steered

Pokémon GO • Feature Technical Lead

Building a Scalable, Data-Driven Biome System

A global case study in turning world data, habitat logic, and authored rules into living encounter spaces.

Research Question

How do you make millions of encounter locations feel local, seasonal, and alive without hand-authoring the world? I led technical development for a biome system that translated map context, ecosystem data, artist-authored rules, and live service configuration into performant mobile encounter environments.

Key Constraints

Constraint

Mobile Hardware

The feature had to hold up on lower-end phones without feeling visually stripped down.

Constraint

Fast Encounters

Catching is quick, so environment setup cost needed to stay low and predictable.

Constraint

Live Service Updates

Seasons, events, and world-state changes needed to ship through data rather than custom engineering work.

Constraint

Location Scale

The system had to respond across cities, coastlines, parks, mountains, and thousands of mixed conditions.

Constraint

Readability

Pokémon could not be obscured, silhouettes had to stay strong, and scenes needed to read instantly.

Constraint

Usability

Artists and designers needed a workflow that stayed usable after the initial feature launch.

Gallery

Role & Scope

  • Led technical development for the feature across engineering and tech art.
  • Designed the Scriptable Object pipeline that maps ecosystem and world data to biome behavior.
  • Built the procedural spawn system for trees, rocks, grass, and supporting environment props.
  • Drove performance work including GPU instancing, shader cost review, caching, and LOD strategy.
  • Built artist-facing tools and workflows so the team could author, preview, and tune content at scale.

Biome Flow

01

World and ecosystem data

Street context, terrain, and environment signals define the input state.

02

Biome rules

Scriptable Object data translates that input into spawn families, overrides, and visual rules.

03

Procedural generation

Seeded placement chooses props and variation while preserving place identity.

04

Runtime rendering

Instancing, LODs, and tuned shaders keep the scene performant on mobile hardware.

05

Live updates

Seasonal, weather, and event data can reshape the same biome without rebuilding the system.

Focus 01

Any Location

A city block, a mountain edge, a riverbank, or open coastline could all produce their own spatial identity.

Focus 02

Persistent Worlds

Variation is seeded to place, so returning to the same location still feels familiar rather than arbitrary.

Focus 03

Season & Weather

Cherry blossoms, fog, holiday dressing, and weather shifts let the same biome keep evolving over time.

Focus 04

Built to Ship

The system was designed to feel rich at a glance while staying fast, readable, and scalable for a live game.

Habitats and Conditions

GrasslandForestBeachRiverLakeOceanCityMountainSeasonalWeatherHoliday EventsHalloween

Procedural Placement

  • Encounter scenes are generated from street, terrain, and ecosystem data instead of hand-authored layouts.
  • Placement is seeded to location so a place can feel persistent rather than random each time you return.
  • Variation happens inside authored rules, which keeps the system expressive without sacrificing readability.

Data-Driven Authoring

  • I designed a Scriptable Object pipeline that maps biome rules directly to spawn behavior and environment configuration.
  • That let the team author new biome variants, tune balance, and support live updates without needing engineering for every change.
  • The result was a maintainable worldbuilding system that scaled better than one-off scene work.

Season, Weather, and Events

  • The same biome can shift across season, weather, and event states, from cherry blossoms to winter dressing to Halloween mood changes.
  • Lighting, fog, and shadow treatments help those changes read immediately without turning the world noisy.
  • This let live service content feel integrated into the system instead of bolted on top of it.

Rendering and Performance

  • I pushed GPU instancing across environment assets, iterated on shader cost, and worked with artists on poly and material budgets.
  • We profiled on lower-end devices, moved nonessential work out of runtime, and cached data wherever it was safe to do so.
  • The goal was quick encounter load time, stable frame cost, and a feature that could survive real production constraints.

Artist Workflow

  • I built tooling that let artists preview, generate, and tune environments instead of fighting raw world data.
  • Rules and overrides gave the team control over composition, special cases, and the tone of each biome family.
  • That balance between authorship and automation is what made the feature feel crafted at global scale.

Findings in the Wild

Shipped as a global live service feature in Pokémon GO.
Supported millions of players across a huge range of world locations.
Gave artists a scalable authoring model instead of one-off scene production.
Created a reusable system for biome, event, weather, and seasonal variations.