Founding Rendering and Perception Engineer
- Menlo Park, CA
- $125k–$250k/yr
- full time
- Posted 1 month ago
About Silicon Dreams
Silicon Dreams is ushering in a future of true abundance for humanity. We are fundamentally rebuilding the way that production and manufacturing is done from the ground up.
Raw material in.
Electricity in.
Human desire in.
Matter out.
We are focused on solving real problems for humans. We believe the most important problem to solve is accelerating a future of abundance across all goods that elevate human quality of life.
About the Role
We are looking for a Rendering and Perception Engineer to own how our robots see — and how we generate what they learn to see. That spans the camera and depth stack on real hardware, the perception models that turn pixels into state, and the rendering pipeline that produces photorealistic synthetic data at the scale training needs. You will be the extreme owner of the pixel path from sensor to policy input, from first camera bring-up to a perception stack running across many deployed units.
This is a generalist role by design. In a given week you might calibrate and time-sync a multi-camera rig, write shaders and materials to make a synthetic scene match a real photograph, train a segmentation or pose model on that synthetic data, chase a perception failure that turns out to be a rolling-shutter artifact, and then go up the stack to build the visualization tooling that makes a failed run legible in thirty seconds. You will have creative freedom to architect from first principles, but you will also own the unglamorous parts: calibration procedures that survive unit-to-unit variation, latency budgets, ground-truth labeling, and the test coverage that keeps a perception regression off the factory floor.
The bar is measured accuracy under real conditions, not a good-looking render. A synthetic dataset counts when a model trained on it transfers to the real camera; a perception model counts when its error is quantified against ground truth across the lighting, occlusion, and material variation the factory actually has.
You need to be a team player, owning all aspects of the end outcomes, as perception here lives at the seam between optics, simulation, machine learning, and the physical machine, and the hard problems are almost always cross-domain.
What You'll Do
Own the real-world camera and depth stack — sensor selection and bring-up, multi-camera calibration, time synchronization, exposure and color pipelines, and latency budgets
Build the rendering pipeline for synthetic data generation: physically based materials, lighting, sensor models, and domain randomization that produces training data which actually transfers
Write and optimize GPU rendering code — rasterization, ray tracing, shaders, and the throughput work that makes million-frame dataset generation affordable
Train and deploy perception models: segmentation, detection, 6-DoF pose
estimation, depth and state estimation for deformable and rigid objects
Own perception evaluation — ground-truth collection, error metrics against real measurements, and failure characterization under occlusion, motion, specularity, and bad lighting
Close the render-to-real gap: measure where synthetic imagery diverges from real captures and fix the renderer or the sensor model, rather than hiding it under augmentation
Deploy perception on real robots at rate — inference latency, edge execution, and graceful behavior when the scene is out of distribution
Build visualization and diagnostic tooling that makes a run, a dataset, or a failure inspectable by anyone on the team
Collaborate closely with the simulation, ML, and firmware work, co-defining the interfaces where perception accuracy is coupled to optics, timing, and scene representation
What You'll Bring
2+ years of experience (engineering club, commercial and research internships included) building real-time rendering, computer vision, or perception systems from scratch
Strong proficiency in Python and C++, and comfort writing GPU code (CUDA, GLSL/HLSL, Vulkan, OpenGL, or a renderer's shading language)
A real grounding in the geometry: camera models, intrinsics and extrinsics, projective geometry, multi-view stereo, and rigid-body transforms — and the discipline to keep frames and conventions straight
Practical computer vision and deep learning experience in PyTorch, applied to perception rather than benchmarks
Hands-on experience with real cameras and depth sensors — calibration, synchronization, distortion, noise, and the failure modes that only show up on hardware
Comfort in a Linux environment, including the host side of a multi-sensor capture system
Strong debugging skill across the optics/software boundary: you can tell whether a bad detection came from the lens, the exposure, the calibration, the renderer, or the model, and you measure rather than guess
Strong first-principles problem-solving and a focus on maintainable, robust code — you care that it keeps working across thousands of runs and many units, not just on your bench
Ability to work independently in a fast-moving, cross-functional environment
If you don’t meet every requirement listed above, but believe you are exceptional, believe in our mission, and are willing to work incredibly hard, please apply regardless. We always want to work with the best.
Preferred Skills
Bachelor's or Master's degree in Computer Science, Electrical Engineering, Robotics, or related
Experience with physically based rendering and production renderers or engines (Blender/Cycles, Omniverse, Unreal, Mitsuba, PBRT, Isaac Sim)
Experience with synthetic data generation at scale and measured sim-to-real transfer results
Experience with neural rendering — NeRF, 3D Gaussian splatting, differentiable rendering, or inverse rendering for material and lighting recovery
6-DoF pose estimation, deformable-object tracking, or state estimation for non-rigid materials
SLAM, visual-inertial odometry, or sensor fusion across cameras, depth, and IMU
Structured light, time-of-flight, or stereo depth pipelines and their error characteristics
Real-time inference optimization — TensorRT, quantization, and meeting a hard frame budget on edge hardware
Color science, HDR imaging, or sensor modelling (noise, rolling shutter, motion blur) at a level that changes what you render
Familiarity with Git, CI/CD, and collaborative graphics or vision workflows
Background in robotics, mechatronics, or physical-systems integration
Compensation
Silicon Dreams is committed to fair and equitable compensation practices. The annual salary range for this role is $125,000 – $250,000. Compensation is determined based on your qualifications and experience. Our total compensation package also includes generous equity in Silicon Dreams.
Working at Silicon Dreams
We believe in having small, talent-dense teams and removing friction across everything. We will always have more problems to solve than people to solve them, and we will keep it this way.
This is how Silicon Dreams builds the best future for humanity in the most efficient manner with a high bar for quality – and this starts with having high quality people who can execute and have a deep desire to be the best in their field.
If you were the one doing all the work in your class projects or college engineering team, you will enjoy working with us.
We index on high gradient and learning rates as opposed to extensive prior experience. We also prioritize candidates with a demonstrated winning spirit whether it is in engineering competitions, physical athletics, or e-sports.
Benefits Summary (USA Full-Time Exempt Employees)
Medical, dental, and vision insurance — no waiting period, coverage effective from your start date.
Commuter benefits, including Caltrain.
Immigration support.
Lunch and dinner provided.
Time Off policies:
3 weeks PTO (15 business days)
5 paid sick days