AI/ML Engineer
Wizard is the AI-native platform for creative work, starting with professional video: a native C++ editor and headless media runtime that humans, agents, and developers all operate.
Own transcription, AI-assisted editing features and local model serving so Wizard's AI works in real time on the desktop.
Why this role matters
Our GTM plan hinges on a five-minute demo: raw footage in, Oz assembles a cut, a finished trailer comes out — and the trailer's quality is the test of whether any of this actually works. You're building the transcription and inference layer underneath that demo: the thing that has to be fast, accurate, and running locally on somebody's desktop, not a cloud API with a spinner.
About the role:
You'll own the transcription pipeline, AI-assisted editing features, and model integration (local and API) that make Wizard's AI features actually work in real time on desktop.
This is one of two AI/ML seats: the first covers multimodal, CLIP-based visual search; this one is focused on transcription/ASR and on-device inference.
What you'll do: Transcription (speech-to-text, waveform text display)
AI-assisted editing features
Model serving (local inference)
What you bring
Must-have experience
Video or audio ML models (not just LLMs)
On-device inference optimization
Whisper or similar ASR systems
Real-time inference pipelines
Show us what you can do
Nice to have: Diffusion models, video understanding
Temporal consistency in video ML
Experience with model quantization / distillation
About Wizard
Every so often, a new technology redefines how stories get told: the printing press, the telephone, radio, the film camera, the cutting room, television, the digital camera, non-linear editing. We think we're at the start of the next one, and we intend to build the company that defines it.
Wizard is the AI-native platform for creative work, starting with the hardest creative workflow: professional video. In roughly six months, our team built a native C++ editor and a headless media runtime that humans, agents, and developers all operate. Oz, our agent, ingests footage at roughly 300× real time and edits at 30–60× real time, and every change it makes stays inspectable, editable, and reversible.
Our philosophy is creative as code: every decision editable, controllable, composable, never generative slop you regenerate and hope for the best. Controllability over convenience. Artist integrity over automation. Taste over throughput.
Our sister company is Story, a production company making video for frontier AI labs and deep-tech companies. We work closely with them: Story is our first customer and in-house design partner, and its editors cut real client work on Wizard every day.
Our team comes from Industrial Light & Magic, Netflix, Apple, Epic Games, Lucasfilm, Blizzard, and 2K. We work in person in San Francisco.