Performance Engineer

Devoted Studios•Montreal, QC

10h•Remote

About The Position

Devoted Studios is a globally remote game development company specializing in Co-development, Porting, and End-to-End Art Production for the global gaming industry. We collaborate across time zones to support projects on all major platforms, engines and styles - from AAA titles to emerging technologies. Our team includes world-class talents who bring deep expertise in external development, pipeline optimization, and creative problem-solving. Whether it’s porting games to new systems, enhancing gameplay features, or crafting stunning visuals, Devoted Studios operates as a trusted, flexible extension of our partners’ internal teams. We are proud to be the development partner of choice for industry leaders such as: Epic Xbox Meta Obsidian Entertainment Riot Gearbox Software In this role, you will own and drive the development of core systems and performance initiatives from planning and design through implementation and optimization. Working closely with engineers and cross-disciplinary teams, you will provide technical leadership and feature ownership.

Requirements

5+ years of professional game engineering experience with at least 3 years in a dedicated performance, engine, or low-level systems role
At least 2 shipped commercial titles where performance was a named ownership area — not a general engineering role that included some profiling
C++ at a deep level: understands memory layout, cache efficiency, branch prediction, SIMD, and how the compiler's optimisation decisions affect runtime behaviour
Unreal Engine 5 profiling expertise: fluent in Unreal Insights, Stat commands, LLM, GPU Visualiser — not just aware of them, but has used them to find real problems on real projects
Cross-platform experience: has shipped on at least two platform types (PC + console, or console + mobile) and understands the meaningfully different constraint profiles
Python proficiency: writes automation scripts, CI integration, data processing for perf telemetry — not a Python developer, but not blocked by Python tasks
Structural fix mindset: profiles, identifies root cause, writes the C++ fix. Does not stop at the recommendation.
Can explain a performance problem to a gameplay engineer or producer without requiring them to understand the GPU pipeline

Nice To Haves

Console certification performance submissions: has prepared and passed Sony/Microsoft performance gates
Mobile performance work on Android (Adreno/Mali/Dimensity) and/or iOS (Apple Silicon GPU) — thermal throttling, tiling architecture, bandwidth constraints
Custom allocator or memory pool implementation in C++
Shader authoring knowledge: HLSL/GLSL fluency sufficient to read a shader and identify unnecessary ALU or texture sample cost
Familiarity with Tracy, Optick, or custom in-house profiling frameworks beyond Unreal's built-in tooling
Experience optimising Nanite, Lumen, or other UE5-specific rendering features at engine depth

Responsibilities

Profile and optimise the game thread, render thread, and RHI thread in Unreal Engine — identifies cross-thread bottlenecks, tick overhead, and task graph inefficiency
Diagnose and resolve hitching: garbage collection pauses, streaming stalls, async loading conflicts, physics simulation spikes
Reduce tick overhead: identifies UObjects and ActorComponents with unnecessary tick enabled, unnecessary work in hot paths, incorrect tick group assignments
Optimise gameplay and engine code at the C++ level — not just profiling and handing recommendations to another engineer; writes and ships the fix
Instruments custom profiling markers (SCOPE_CYCLE_COUNTER, CSV_SCOPED_TIMING_STAT, custom Unreal Insights channels) for ongoing regression detection
Profiles rendering performance using RenderDoc, PIX, NSight, Xcode GPU Frame Capture, and Unreal's GPU Visualiser — isolates draw call overhead, overdraw, shader complexity, and GPU memory bandwidth
Works with the rendering team to optimise material pipelines: shader permutation reduction, LOD tuning, occlusion culling, distance field shadow optimisation
Identifies and resolves GPU memory pressure: texture streaming configuration, render target pooling, VRAM budget management per platform
Understands the rendering architecture deeply enough to distinguish CPU-bound vs GPU-bound vs bandwidth-bound frames and route the fix correctly
Targets platform-specific GPU constraints: the PS5/Xbox Series GPU architecture differs meaningfully from PC; mobile (Adreno, Mali, Apple GPU) has its own thermal and bandwidth profile
Owns memory budget compliance on client engagements: tracks resident set size, texture memory, audio memory, and engine overhead against platform limits
Finds and eliminates memory leaks in C++ using both automated tooling (ASan, Valgrind on non-console platforms, custom allocator hooks) and Unreal's LLM (Low-Level Memory tracker)
Manages memory fragmentation: understands pooled allocators, custom memory arenas, and when Unreal's default allocation strategy creates fragmentation under long session times
Console memory work is the hardest version of this: fixed RAM ceilings with no swap. Has shipped inside those constraints.
Builds memory regression tests that run in CI — not just point-in-time snapshots
Profiles and optimises Unreal's replication system: identifies bandwidth-heavy actors and components, tunes replication frequency, prioritises replication correctly under load
Diagnoses client-server latency: distinguishes network latency from server frame time from client simulation cost — does not conflate them
Familiar with NetStat, Unreal's network profiler, and packet-level capture when needed
Supports multiplayer engagements where networking is a performance constraint — not the primary discipline on pure-networking architecture
Builds internal profiling tools in C++ (UE plugins) and Python — automated perf validation scripts, CI-integrated regression detection, custom Unreal Insights data channels
Writes automated performance regression tests: baseline capture, threshold alerting, trend visualisation — so regressions are caught in the build, not in the client review
Builds or improves developer-facing profiling UX: makes it easier for other engineers on the engagement to find their own bottlenecks without specialist involvement
Documents profiling methodology and tool usage: other engineers can reproduce the analysis without a knowledge transfer session