Mirage is an AI video company focused on making creation dramatically easier. Our team tackles some of the hardest creative and technical challenges in generative media.
We get to rethink how people collaborate with AI that can actually do creative work with them—defining entirely new interaction paradigms for AI-native video tools. We make that possible by building multimodal agents that turn creative direction into edits, rendering engines that composite multiple layers of video and graphics, and generative video models trained from the ground up that create new media when needed.
All of this comes together in Captions, our creative workspace for generating, editing, and designing video. The technology behind Captions is now available more broadly: Tesseract gives any AI agent native video-editing capabilities, while developers can integrate our models into their own products through APIs.
Explore our work Captions — Our flagship creative workspace Tesseract — Our professional video engine for AI agents Research — The foundation models we build in-house Updates — The latest from Mirage TechCrunch, Forbes, Fast Company — Press
Our Investors We’re very fortunate to have some the best investors and entrepreneurs backing us, including Index Ventures, Kleiner Perkins, Sequoia Capital, Andreessen Horowitz, General Catalyst, Uncommon Projects, Kevin Systrom, Mike Krieger, Lenny Rachitsky, Antoine Martin, Julie Zhuo, Ben Rubin, Jaren Glover, SVAngel, 20VC, Ludlow Ventures, Chapter One, and more. Please note that most of our roles will require you to be in-person at our NYC HQ (located in Union Square)
About the role We want Tesseract to render 3D that is indistinguishable from footage. You'll be the founding engineer for photoreal 3D in Tesseract, our Rust-based, cross-platform graphics and media rendering engine. Tesseract already renders video, motion graphics, effects and compositing in a color- managed pipeline. You'll build 3D to the same standard: a fast interactive preview for editing, and a physically based final render, path-traced or hybrid ray-traced, that produces film-quality frames for export. Both run from one scene, material and lighting model, so what people see while they edit is what they export. Scenes are authored by people and by AI agents, so render cost, failure modes and APIs matter as much as image quality. This is a foundational role: you'll make the core architectural decisions and help build the team around them. What you'll do
- Define the architecture for photoreal 3D: how ray tracing, rasterization and denoising divide the work, and how that maps onto Metal, Vulkan, Direct3D and WebGPU.
- Build light transport: global illumination, reflections and refraction, soft shadows from area lights, and volumetrics such as haze, fog and light shafts.
- Build a physically based model for materials, lights and cameras: energy-conserving materials (metal, glass, clearcoat, sheen, translucency), lights in real-world units, and a camera with exposure, depth of field and motion blur.
- Own image quality: sampling, antialiasing, accumulation, denoising and firefly suppression, with frames that stay stable and deterministic across a video export.
- Scale scenes to production complexity with acceleration structures, instancing, culling and memory management, for millions of triangles and tens of thousands of objects.
- Integrate 3D into Tesseract's video, motion graphics and compositing pipeline, including ACES color management and render passes such as depth, normals, motion vectors and light groups.
- Make render cost predictable and rendering robust: performance budgets, profiling, and headless batch rendering on local and cloud GPUs.
- Design 3D APIs that work for people and AI agents alike: a clear scene representation, actionable validation errors, fast previews and render cost known before the render.
- Work with engine, product and AI teams, and help hire and grow the 3D rendering team.
What makes you a great fit
- Deep experience building physically based renderers in production: film or offline renderers, path tracers, or game engines with ray-traced or hybrid lighting.
- Strong command of light transport: the rendering equation, Monte Carlo integration, importance sampling, material models and denoising.
- Hands-on GPU ray tracing: building and traversing acceleration structures (BVHs), hardware ray-tracing APIs such as Metal ray tracing, Vulkan ray tracing or DXR, and compute-based fallbacks.
- A solid understanding of modern rasterization pipelines, and of how hybrid renderers combine rasterization with ray tracing.
- Hands-on experience with Metal, Vulkan, Direct3D or WebGPU, including compute shaders, GPU profiling and GPU debugging.
- Proficiency in Rust or C++, with strong reasoning about memory, concurrency and performance.
- Strong foundations in the mathematics of graphics: linear algebra, geometry, probability and sampling.
- An eye for realism: you can look at a frame, say why it doesn't read as real, and fix the cause.
- Experience owning complex technical projects and making architectural decisions in ambiguous problem spaces.
Even better if…
- You've worked on the internals of a production renderer or engine such as Cycles, Arnold, RenderMan, V-Ray, Karma, Unreal Engine (Lumen or its path tracer) or Unity HDRP.
- You've built or integrated ML denoisers, or worked on neural rendering.
- You've worked with color management (ACES, OpenColorIO), HDR output or film-look pipelines.
- You've built on cross-platform graphics abstractions such as wgpu, including extending them where they lack features such as ray tracing.
- You've optimized rendering for Apple silicon or mobile GPUs.
- You have experience with Rust, video compositing, or integrating 2D and 3D rendering in creative tools.
Benefits
- Comprehensive medical, dental, and vision plans
- Catered lunch multiple days per wee
Sourced from the a16z Speedrun talent network — apply on Speedrun.