I placed 2nd in the Ubisoft Toronto NEXT Level Design Competition 2026. This project is a re-imagining of Marineland as a compact stealth map with recognizable landmarks, clear routes, and lots of small decisions that make players feel sneaky as they outsmart enemies.
Role: Level Designer
Project length: Sep 8, 2025 - Feb 11, 2026
Tools: Unreal Engine, Illustrator, Blueprints
Game: Ubisoft Toronto NEXT Competition - Level Design
Placement: 2nd Place Finalist
What I wanted to achieve with the level
I wanted to build a stealth level where players always have a plan, even if that plan changes mid-run.
Design intentions
Support multiple playstyles and paths
Keep the park feeling realistic, dense, and readable, with clear landmarks and quick decision loops
Deliver memorable traversal payoff using the Sky Drop tower as a set piece
View of Full Level
How I wanted to achieve it
I built the level around “choice moments” and made sure each beat had at least two valid answers.
My approach
Bake options into the main beats (front gate, theatre, aquarium), so routes are not “bonus paths”, they are real solutions
Use readability tools to guide players without UI hand-holding (leading lines, cables, sightline breaks, hard locks)
Teach a new mechanic in a safe space, then re-use it later under pressure (Dry Ice fog)
Constraints I designed around
These rules actually helped the level get sharper and more intentional.
The challenge required a linear stealth action-adventure level with traversal challenges, light hostile presence, and a required lock-and-key
Production limits forced smarter layout choices: 12 AI cap and a condensed play space
Process
Phase 1: Mission Design Document (top-down)
I started by mining the instructions for constraints and “must-haves” (lock-and-key, traversal, multiple solutions) , then brainstormed grounded real-world locations. Marineland clicked because it naturally gives you big landmarks and clear themed zones.
I learned Illustrator during the competition, and honestly, tracing patrol loops with my finger on the screen helped me find routes that were both readable and interesting.
Midpoint check-in
I used the check-in to validate three things: whether the level still encouraged stealth, whether sightlines felt fair, and whether the beat-to-beat flow stayed clear even without objectives fully implemented yet .
Phase 2: Playable blockout
When I moved into Unreal, I had to tighten scope fast: shrink the footprint, reduce enemy count, and clean up areas that were cluttered or hard to read .
Flow Chart of Beats
Obstacle placement
Enemy Placement
What changed and why
1) Exterior readability and guidance
Problem: the outside space felt too open and objectives were easy to drift past.
Change: condensed building scale, added key in fountain objective, and used cables/leading lines to connect related interactions .
Result: players had stronger direction without losing freedom.
2) Theatre clarity and stealth options
Problem: theatre space was cluttered and hard to parse.
Change: restructured to be less congested, improved sightlines/obstacles
Result: cleaner pathing, better scouting, and can learn the dry ice mechanic
3) Better Sky Drop Payoff
Problem: riding the tower was functional but not exciting.
Feedback: “Tower is boring for something so special looking.”
Change: turned it into a sequenced event where it breaks halfway, forcing a climb with a real traversal finish.
Result: the tower became a memorable story beat, not just an elevator
New ingredient: Dry Ice Smoke
I wanted a stealth tool that fit the location and solved a real level problem: long, monitored sightlines inside Arctic Cave. Dry ice already existed in the fiction as show fog, so repurposing it into a throwable smoke tool felt grounded and “in-world,” not a random gadget.
Design goal: give players a way to create temporary cover where none exists, without turning it into a shoot-out.
What it does for stealth
Breaks sightlines in tight corridors and observation halls, letting players cross “impossible” spaces without defaulting to kills.
Creates safe lanes through patrols and cameras by shrinking detection while the fog is active.
Enables a clean, non-lethal interaction with the VIP courier by masking the pickup moment.
Blueprint implementation
I built it as a small chain of simple Blueprints that plug into existing project patterns rather than rewriting core systems.
Pickup + UI
BP_DryIcePickup uses the same structure as ammo pickups: collect, hide the mesh, add to the player, update UI.
BP_PlayerCharacter increments the dry-ice count and updates the HUD.
Deploy
BP_DryIceDeploy activates particles on impact, enables a collision/overlap volume, and checks enemies inside the cloud to trigger a behavior change event.
Throwing uses the same baseline logic as the existing rock throw, just tuned for this item.
AI response
I handle the “lost in smoke” effect by clearing player-location related values and reducing sight while inside the radius, then restoring normal behavior when exiting.
Playtesters had trouble following button connections. I used cables and leading lines to guide players.
Front gate ladder
Playtesting
I ran structured playtests with specific questions around ammo balance, readability, enemy path clarity, objective confusion, and route choices.
My workflow:
Collect feedback
Sort into “can act on” vs “cannot act on”
Prioritize by effort vs impact
Patch, then retest
Multi-route stealth
Theatre: Castle
Sky Drop to Aquarium
What I learned
Constraints are not a limiter, they are a design tool. The AI cap and space limits forced me to cut “filler” and keep only meaningful guards and cameras .
Teaching a new mechanic needs a deliberate “teach → test → challenge” plan. I got close, but I would structure it even earlier next time .
A big landmark must earn its screen time. Turning the tower into a traversal sequence was the single highest-impact change.