Careers

Optra builds Optra Studio and OptraCloud. AI agents write most of our code, so every person who joins raises the ceiling of what we can build.

Two roles are open. Each one is described below, with how to apply.

Fit large open models onto consumer GPUs, and fine-tune our inference. At Indian prices, efficiency is our margin.

Get large open models to run well on the consumer GPUs we have: choose how to quantize them, how to batch requests and how to serve them.
Fine-tune our inference: measure what each answer costs us in GPU time, change one setting at a time, and keep the change that brings the cost down.
Keep what you build running for the people who use it, and fix it when it slips.

You have served a large open model yourself, on a consumer GPU, and can say what it cost to run.
You know quantization, batching and a serving stack such as vLLM from using them, not only from reading about them.
You measure before you claim, and you can show a result.

Email [email protected] with your CV and one thing you built. Put the name of the role in the subject.

Research video and audio models, and fine-tune the ones we use. This is what makes our films and voices better.

Read the new research on video and audio models, test what is worth using on the hardware we have, and say plainly what is not.
Fine-tune the models we choose, so that our clips and voices improve, and show the change beside the version before it.
Judge the output by looking and listening, not only by a score.

You have fine-tuned a video, image or audio model yourself, for example with a LoRA, and can show what changed in its output.
You read research papers and can tell which ideas will work on a consumer GPU.
You compare a change with its baseline before you call it better.

Email [email protected] with your CV and one thing you built. Put the name of the role in the subject.