Cast your mind back to that closing second pass I mentioned. This is the part that settles the call, the last
gate before an interview. The recruiter eases off and reads more closely here, yet
95% of the screen still rides on your most recent role all the same.
That stands to reason: your latest role is the sharpest read on where your seniority lands today, what you
can do, and what really fills your day. To win the "yes", that role needs to step through the
full role profile for a Graphics Engineer, one dedicated bullet per area that you
already flagged under Domain Expertise back in the Profile Summary.
1
Render Pipeline Architecture
You architect how every frame gets drawn. The pipeline shape sets the ceiling on everything above it, so
hiring managers want a render architecture you designed, not inherited. Talk about how you used a
clustered pipeline and a frame graph, over Vulkan and DX12, to cut GPU frame time and pass cost.
Techniques
Forward / deferred / clustered
Frame graph & render passes
Render thread & sync
Barriers & resource state
Tools
C++17/20, HLSL/GLSL
Unreal source, in-house engine
Vulkan, DX12
Metrics
GPU frame time (ms)
Pass cost (ms)
Draw calls
2
Shaders & Materials
You write the shaders that make surfaces look right. Shader cost shows up straight in the frame budget,
so hiring managers want shader cost and variant counts you held. Show them how you used PBR materials
and permutation control, in HLSL and Slang, to cut shader cost and variant count.
Techniques
PBR materials
Shader graph & nodes
Permutations & variants
Compute prepasses
Tools
HLSL, GLSL, Slang
DXC, shader compiler
Material editor
Metrics
Shader cost (ms / cycles)
Variant count
Compile time
3
Lighting & Global Illumination
You light the scene in real time. Real-time GI is one of the hardest problems in graphics, so hiring
managers want lighting cost you actually held. Point out how you used real-time GI and clustered light
culling, with a Lumen-style approach, to hold lighting cost at high light counts.
Techniques
Realtime vs baked GI
Shadow maps & cascades
Reflections & probes
Light culling
Tools
Lumen-style GI, lightmaps
Tiled / clustered culling
Ray tracing, SSR
Metrics
Lighting cost (ms)
Light count, shadow res
4
Post-Processing & Image Quality
You take the raw frame and make it look finished. Hiring managers look here to see whether you lift
image quality within budget, or blow the frame time chasing pixels. Mention how you used tone mapping
and TAA with DLSS upscaling, in compute shaders, to hold post cost while lifting image quality.
Techniques
Tone mapping & color grading
TAA & anti-aliasing
Upscaling (DLSS, FSR)
Bloom & depth of field
Tools
HLSL / GLSL, compute
DLSS, FSR, XeSS SDKs
RenderDoc, PIX
Metrics
Post cost (ms)
Output resolution
Image quality delta
5
GPU Performance & Profiling
You find what's choking the GPU and fix it. A high average FPS hides GPU stalls, so hiring managers
read occupancy and bandwidth work as real GPU optimization. Walk them through how you used occupancy and
bandwidth analysis, captured in RenderDoc and Nsight, to cut GPU frame time and lift occupancy.
Techniques
Occupancy & bandwidth
Draw-call batching
Bottleneck analysis
GPU capture & trace
Tools
RenderDoc, PIX
Nsight Graphics
Radeon GPU Profiler
Metrics
GPU frame time (ms)
Bandwidth (GB/s)
Occupancy %
6
Graphics APIs & GPU Programming
You program the GPU directly through modern APIs. Two things ride on it for a hiring manager: draw calls
cut by GPU-driven rendering, and dispatch cost held with bindless. Lay out how you used GPU-driven
rendering and bindless descriptors, in Vulkan and Metal, to cut draw calls and dispatch cost.
Techniques
RHI / API abstraction
Compute shaders
GPU-driven rendering
Bindless & descriptors
Tools
Vulkan, DirectX 12, Metal
HLSL / GLSL, DXC
Validation layers
Metrics
GPU frame time (ms)
Draw calls
Dispatch cost
7
Geometry, LOD & Culling
You push only the triangles that matter to the GPU. Wasted geometry burns the whole budget, so
virtual-geometry work tells a hiring manager you optimize where it counts. Spell out how you used
virtual geometry and GPU occlusion culling, with mesh shaders and Hi-Z, to cut triangle count and VRAM.
Techniques
Mesh pipeline
Virtual geometry (Nanite-style)
Instancing
Frustum & occlusion culling
Tools
Mesh shaders
GPU culling, Hi-Z
LOD & impostor tooling
Metrics
Triangle count
Draw calls
VRAM (MB)
GPU frame time (ms)
8
Tooling, Debugging & Cross-Platform
You debug the GPU across every platform. Companies keep the graphics engineers who ship visual parity
everywhere, not the ones who tune for one GPU, so this stands out. Tell them how you used GPU capture
and shader debugging, in RenderDoc and PIX, to ship visual parity across platforms.
Techniques
GPU capture & debugging
Shader debugging
Validation layers
Cross-platform parity
Tools
RenderDoc, PIX, Nsight
Console GPU tools
Vulkan validation, GPU markers
Metrics
Platforms shipped
Visual parity
Repro time