Published by 10xCareer Community. Read this scan, then run one on a job you are weighing.
Video Algorithms Intern, Video Coding (Gaussian Splatting), Fall 2026
Netflix · Los Gatos,California,United States of America
This posting welds three distinct roles into one: 3D computer vision research scientist, video codec engineer, and graphics software prototyper.
At 30% automatable, AI can speed up boilerplate PyTorch code and hyperparameter sweeps. This number misleads because the core mandate is frontier research into unsolved 3D rendering compression problems where no pre-existing AI solution exists.
Automation exposure
A small slice of this job is automatable today.
The rest is where the role actually lives — and where you come in.
Where you'd be irreplaceable
The part of this job AI can't touch — and your line for the interview.
“The part of my job AI can't touch is inventing novel compression algorithms for 4D Gaussian Splatting to make photorealistic 3D streaming feasible on consumer hardware.”
What the whole role is built around
Inventing and validating novel compression and fast-training algorithms for Gaussian Splatting that reduce model sizes down to commercial streaming bitrates.
Where you'd add value a tool can't:
- Formulating original algorithmic approaches to compress 3D Gaussian representations.
- Evaluating visual perceptual quality artifacts that standard automated metrics miss.
- Architecting end-to-end proof-of-concept renderers for consumer playback hardware.
Neural rendering and 3D Gaussian Splatting will replace traditional video codecs like HEVC and AV1 for immersive media within a decade.
What you'd actually do all day
Every duty in the posting, and who's better placed to do it today.
- Explore GS model compression strategiesYou + AI
AI generates standard pruning scripts, but theoretical strategy selection needs researcher intuition.
- Characterize trade-offs among GS model sizeYou + AI
Automated benchmarking runs metrics, but setting perceptual target balances requires human judgment.
- Identify strategies to reduce training timeYour edge
Creating novel research algorithms for 3D scene reconstruction requires original human innovation.
- Design and implement a proof-of-conceptYou + AI
LLMs draft code scaffolding, but custom graphics pipeline integration requires manual work.
- Contribute to dataset needs for scenesYour edge
Selecting representative visual scenes requires understanding product edge cases and streaming constraints.
How the role scores
Three quick reads, each out of 10. Tap “Why” for the reasoning.
Where the field is heading
Growing · 10+ years
Demand for spatial computing and 3D/4D photorealistic rendering is accelerating across hardware vendors and streaming platforms. Efficient compression of novel-view synthesis models is essential for bandwidth-constrained consumer playback.
Your next moves
Where to start if you want the edge this role rewards.
Learn these
Direct GPU programming expertise significantly increases your market value and research output speed in neural graphics.
Combining traditional video coding principles with 3D computer vision makes you uniquely qualified for senior streaming algorithm roles.
Understanding visual quality evaluation beyond PSNR/SSIM is essential for productionizing real-time graphics.
Try these tools
- GitHub Copilot
- Writing PyTorch model wrappers and evaluation boilerplate for Gaussian Splatting
- Weights & Biases
- Tracking metrics when characterizing trade-offs among GS model size, training time, and rendered quality
- Claude 3.5 Sonnet
- Refactoring complex CUDA memory management code and generating mathematical derivations
Ask about this before you apply
Things the posting never says that would change the picture.
- Specific hourly pay rate for this exact role within the listed broad internship range ($40-$110/hr)
- Specific hardware target specs (e.g., target bitrate, memory limits for TV streaming sticks vs phones)
Hiring for a role like this? See how teams use scans to write clearer job posts.