Oliver Neal

In development · Unity 6

Out of Water

A turtle and a seagull, in splitscreen or online.

The shot is the team's art. My side is the tech underneath it.

Shiverbug Studios Middlesbrough, UK

1. Technical Director (my page on the studio's site) Oct 2026 to now · board
2. Gameplay Programmer Intern Jul to Sep 2026 · 10 weeks

• Lead the tech Engineering on the game, and the technical plan for the vertical slice we're building now. The rest of the team are artists, designers and animators, so most of the code comes through me.
• Run the pipeline I'm the studio's buildmaster: the git workflow, the repositories and the playtest builds. The second decision below is how it works.
• Unblock the team Artists and designers working confidently in Unity and git, so an art-only change never has to wait for me.

1. Online co-op

   Each player's machine owns their own character

   Online play is client-host. Your machine runs your character exactly as it does in splitscreen, your partner's character is a puppet on your screen, and the host owns the world: enemies, puzzles, pickups and the save.

   Why

   A platformer lives on input latency, and this way your own character never waits on the network. The state machines that already drive both heroes run untouched. Rollback was the alternative, and it needs a deterministic simulation, which this game does not have.

   Cost

   The two players can disagree for one round trip. That only matters where the characters touch each other, so those moves need the most care.

   Tested

   With two instances at a simulated 150 ms of lag, before it reaches a playtest. How it feels on a real connection is still to be proven.

2. The pipeline

   Every change lands through a pull request

   Everyone works on their own branch and opens a pull request into a shared integration branch. Each one goes through me or one of the founders. A small change is reviewed as a diff; a bigger one is opened in Unity and played before it merges.

   Why

   A Unity merge can come back with no conflicts and still be wrong. Scenes, prefabs and materials do not merge as text, so no conflicts means the text lined up, not that the project still works.

   Cost

   Finished work waits on a reviewer. And my own code is the gap: I merge my own pull requests, so it leans on the automated tests and the playtest build, not on a second reader.

   Result

   A tagged playtest build every fortnight. Anyone who would rather not open a pull request pushes their branch and tells me, and I open it for them.

3. Architecture

   Layers that only point one way

   The code is split into core, game, online and UI layers. Each is its own assembly, so the compiler refuses a reference that points the wrong way. Logic that needs no engine sits at the bottom, where it can be unit tested.

   Why

   Gameplay code named the online session in 1,022 places, by the count of the repo's own metrics tool, so a feature and its netcode could only ever change together.

   Cost

   It is not finished. Inside the game layer the player, the enemies and the level tools still reference each other, and that is the next cut.

   Result

   That count is now zero. Automated tests hold the feel of the moveset and the combat to recorded baselines.

4. Performance

   Nothing is faster until it is measured

   Performance work is compared branch against branch, on the same scene at the same quality tier, with a profiling tool I made for the team.

   Why

   Split screen draws the world twice, so a change that looks free on one camera can cost double. A number settles the argument.

   Cost

   It is slower. A change is not called faster until the same scene has been captured on both branches.

   Result

   The latest pass took 15 to 18% of the draw calls out of split screen on the lowest quality tier. What I do not have yet is a frame time for the whole game on the hardware we are aiming at. That benchmark is still to run.

Apr 2026 · Unity 3D · Solo

Cyber Station

A neon station management game about building a transport hub that can take the pressure. 86 scripts and around 17,000 lines, on my own.

Hundreds of passengers run off one fixed 10 Hz tick, and each picks a facility by lowest estimated total delay: queue wait plus real NavMesh walk time, both in seconds.

Read the breakdown Play Source

May 2025 · Unity 6 · Solo · Game jam

Wanted

A top-down stealth survival game with no win state, only a clock. The police arrive car by car and never leave.

One detection meter fills four times slower under cover, and a confirmed sighting redirects only the closest 10% of officers, so the response scales with the force.

Read the breakdown Play Source

Aug 2025 · Unity 3D · Solo

Minimalists

A node conquest RTS pared back to its essentials: capture structures, commit forces by percentage, and win anything from a duel to a four way free for all.

The enemy AI has no extra information. It scores every source and target pair by distance and garrison, then commits through the same send call the player's slider drives. There are six of them.

Read the breakdown Play Source

Feb 2025 · Unity · Solo

Unicellular

An idle simulation that rewards the player for pushing the entity count up, so every cell has to stay cheap to simulate.

Stress-tested at 10,000 cells, still reading around 90 FPS in the recording, on a multi-rate tick with the proximity checks time-sliced across frames.

Read the breakdown Play Source

More work

• Tile Turfer

  A turf-war shooter where one territory percentage drives difficulty, fire rate and enemy AI at once, across procedurally generated caves.

  Jan 2026 · Unity 2D · Solo

• Space Bar Simulator

  A first-person bar sim built by a team of ten. I owned the player controller, the interaction system and the camera-blended stations.

  Jan 2026 · Unreal Engine 5 · Team of 10

• 3D Netcode Tank Game

  Server-authoritative multiplayer on Netcode for GameObjects: Relay sessions, and a 64-ray hitscan resolved on the server.

  Jan 2025 · Unity 3D · Solo

• Pinnable

  An arcade catch game with a spawn director that scales each hazard's weight with your score.

  Mar 2024 · Unity 2D · Solo

• Candy Corporate

  A top-down extraction game made with a small team. I did the level design, tile creation and pixel art integration.

  Apr 2024 · Unity 2D · Team · itch.io

What I work with

Languages & engines

C#, Unity 6 (2D / 3D), Shader Graph, HLSL, Unreal Engine 5, Python

Gameplay & AI

Gameplay programming, systems design, finite & hierarchical state machines, AI navigation, local co-op & splitscreen, Unity Input System

Performance

Profiling and frame analysis, time-sliced & multi-rate simulation, GPU-driven rendering, LOD and culling, quality tiers, runtime NavMesh baking

Multiplayer & services

Netcode for GameObjects, Unity Authentication, Relay, Cloud Save

Audio, UI & presentation

FMOD, UI Toolkit, Unity UI, Cinemachine, particle VFX

Tools & workflow

Git and Git LFS, branch workflow and pull request review on a shared team repo, automated testing (Unity Test Framework), Visual Studio, JetBrains Rider, Figma, Trello, technical documentation