Machine Learning Research Engineer

At Phonic, we are building a platform to help users build, observable, and evaluate voice apps with a focus on making them reliable. Phonic helps the user increase reliability of voice agents while offering lower cost and latencies. 

To do so, we are training audio foundation models from scratch on petabytes of data that speak with life-like conversationalness at extremely fast latencies. We are a team of experts from MIT and Stanford working on cutting-edge model research and ML infrastructure scaling challenges to unlock new generative capabilities that are a step beyond what is possible today. We've raised a seed round from Lux Capital and are looking for a Machine Learning Research Engineer to join our office in San Francisco in our mission to reinvent the future of audio generation.

 Some potential areas that you could work on:

• Training Runtime: you'll build fast, cloud-native training infrastructure that minimizes the time for a job to launch, a robust job orchestrator, deterministic streaming dataloaders, and optimized model implementations.
• Evaluations: you'll help evaluate our models on automated metrics such as quality and word error rate, and subjective metrics like an ELO leaderboard for human annotators.
• Inference: you will architect, build, and deploy the backend systems and services that power our audio foundation models. If you nerd out over things like continuous batching, minimizing time to first byte, and batching requests, this might be a good fit.

*There are no hard requirements, as long as you can demonstrate you are able to write a lot of good code*

• Previous experience working with big data pipelines and infrastructure to support large-scale model training.
• Ability to learn and iterate quickly.
• Self-motivated with a willingness to take ownership of tasks.
• Take pride in building and operating scalable, reliable, secure systems.
• Own problems end-to-end, and are willing to pick up whatever knowledge you're missing to get the job done.