Senior Machine Learning Engineer

About us

Our mission at Speak is to teach the next billion people how to speak English.

We believe that automatic speech recognition and machine learning will transform how people learn English. We're building this future: software that can understand what a student is saying, give them immediate feedback, and guide them towards fluency, all on their phone.

Currently, almost nobody learning English approaches fluency because doing so is expensive, arduous, often embarrassing, and requires human tutoring at hourly rates. We believe that by using AI and technology to replace much of the human involvement in the English learning process, we'll be able to help anyone that wants to learn this life-changing skill.

Speak is one of the fastest growing language learning apps in the world - we're growing 30% MoM and our users generate over 10,000 minutes of training data every day.

About this role

We're looking for a senior ML engineer to take a central role on Speak's machine learning & speech team. This is a very special opportunity for the right engineer to work on enormously impactful problems with clear solutions powered by real-world deep learning. You'd have complete ownership over our machine learning projects and work extremely closely with product, engineering, and design, to ship custom models from conception through deployment to millions of users.

The right candidate will be obsessed with and experienced in building and training custom models, have a strong nose and developed intuition for continually making progress, and be willing to do whatever it takes to make the model work in production.

What you'll do

• Build custom ASR, speech, and audio models from conception to training to deployment in production. Think custom small-vocabulary ASR model, sequence-to-sequence phoneme recognizer, or audio style transfer model.
• Collect, structure, and clean training data at scale. Think accented speech from our users, academic datasets, and crawling the web for speech audio data.
• Work closely with almost every other team at the company (product, engineering, design, content) to build user experiences and productize these models

What we're looking for

• You've written, trained, debugged, and deployed several custom models, and can tell the story around how and why you did it, how you solved issues and bugs, and how you created a meaningful end-user experience.
• You love getting your hands dirty, building models and knocking down roadblocks, and trying out new techniques for fun.
• Ideally, experience with acoustic modeling and speech/audio data, or other real-time signal processing.

Why work on ML at Speak?

1. Join a fantastic, tight-knit team at the right time: we're growing super quickly - our subscriber base has grown 20x in the past 12 months, and we've achieved product-market fit. You'd join at a magical time when a single person could significantly change the course of the company. We're also backed by some of the best startup and AI investors in the world.
2. Innovate with a huge, exciting, and unique dataset: we have a giant dataset of English speech data that's growing every day, and a large set of problems that we're trying to solve. This is an opportunity to work with an almost unlimited amount of valuable data and build out models that could directly transform people's lives.
3. Immediately impact the product: there aren't many consumer products where machine learning actually occupies a central role in the product experience; Speak is one of them. You'd be applying and honing your ML skills to build models that directly power core product experience; get something working well and it could be in the app in a week.
4. Ownership & responsibility: this role offers a lot of autonomy, responsibility, and ownership within Speak's ML efforts. Since we're so small, you'll work with almost every other team. This means you'll have both the autonomy to try out new ideas and the resources to make a product vision a reality.