We’re creating Voice for everyone.
At WellSaid Labs, we enable creatives around the globe by putting high-tech, human parity technology into their hands, giving them the ability to add voice-over to any project and iterate with ease. Creative teams use WellSaid Lab’s Voice Studio to create compelling employee training, design unique digital experiences, and narrate audiobooks. We believe deeply in AI for Good, and that technology should be empowering, engaging, and fair to all people.
Who You Are: a Data-Driven Applied ML Engineer, working in Text-to-Speech
The WellSaid Labs Applied Machine Learning Team works with stakeholders to identify, refine, and solve problems at the intersection of machine learning and customer needs. This team understands customer needs through quantitative and qualitative research and they work across WSL teams to understand how machine learning is utilized and how it can be improved.
Improving and maintaining our ML solutions includes creating test datasets and metrics to define and gauge success, working with the ML Platform Team to prioritize model updates, training new models for deployment, coordinating releases, and educating the customer on any new capabilities. The Applied ML Team is consistently testing and reviewing any deployed models.
As a Text-to-Speech Data Engineer on the Applied ML Team at WellSaid Labs, you will be working to regularly improve our text-to-speech service. You will add new datasets; train, deploy, and evaluate new models; and design experiments and algorithms for solving new and creative TTS challenges. You will report your findings to the Applied ML Team who is then actioning any improvements. It would be great for you to be familiar with language construction, voice over and/or audio processing. You may also work closely with our Voice Audio Engineer in establishing data requirements, assessing audio, and generating new script content for voice talent to read.
How You’ll Contribute:
In your day-to-day, you will:
- Work directly with text and audio data: gathering, compiling, and organizing datasets, preparing data for machine training, evaluating results, debugging problematic data
- Train and deploy ML models: incorporating new data, monitoring training metrics, debugging failing code, deploying a model for customer use
- Evaluate ML models: consider causation or correlation between training data and ML predictions, design ML experiments and establish success criteria, gather and evaluate metrics including mean opinion scores
- Additional research projects: interesting data or use cases, alternative services and solutions, internal process improvements, new data evaluation tools, etc
- Create Datasets: new voice-over scripts, new recordings, new delivery styles, test datasets
- Evaluate ML Models: pronunciation accuracy, naturalness, text normalization coverage, loudness accuracy, speed accuracy, language and dialect expansion, customer-model friction
- Train/Deploy ML Models: marketplace releases, custom voices
- Monitor Usage: identify model production errors, beta program usage and feedback
- Research: speaker blending, text verbalization, new languages and dialects, new TTS features, new STT services
Additionally, this role requires you to write and execute code that enables you, and others, to perform each of these tasks. It also requires you to think critically about language, dialect, pronunciation, phonemics, and audio dynamics in order to build the highest quality voices and Studio/API experiences for our customers.
What We’re Looking For
To thrive in this role, you ideally have experience with and a solid understanding of ML concepts and best practices, successfully managed datasets and metrics in a ML capacity, coding experience developing tools that can evaluate data and enable you to establish recommendations based on data analysis results, and experience with software releases to production.
- You have worked in a technical team, managing project expectations and communicating your plans, project statuses, and results frequently and comprehensively for 3-5 years.
- You have worked with a wide array of data types, building analysis tools and establishing success criteria for evaluating the success of data-driven projects
- You have built and deployed ML models for use by a non-technical audience, clearly communicating usage guidelines and best practices
- You have experience building and documenting new processes, especially in a ML pipeline or similar capacity
- You have a strong understanding of the importance of data preparation for ML training, data visualization and metrics for ML assessment, and analysis of ML results
- You have familiarity with software and feature releases and can work closely with a Product team for exposing ML changes to customers
- [Bonus] You have a curiosity and interest in linguistics and acoustics
- [Bonus] You have an interest in eventually managing an Applied ML Team, strategizing workload, and mentoring contributing engineers under your supervision
- [Bonus] You have studied Deep Learning and have applied models to solve technical challenges
To join our team you must also:
- be a U.S. Citizen or Permanent Resident
- pass a pre-employment background check
What We Offer
WSL is proud to support an inclusive work environment that emphasizes each team member’s personal and professional growth. Our team is fully distributed throughout the U.S., and we support flexible schedules – work where and when you work best. You’ll have teammates just a Slack message or video call away if you ever need help solving an exciting challenge, or even if you just have a funny story to tell.
Other perks and benefits:
- Competitive salary and stock options
- Full medical, dental, and vision insurance
- Matching 401(k) plan
- Generous vacation policy/paid time off
- Parental leave
- Learning & development stipend
- Home office stipend
As a startup, we strive to be externally competitive with companies at a similar size and stage, and internally fair in our pay practices. The hiring salary range for this role is $130,000 – $160,000, and represents the target offer range given the scope and experience expectations for this role.
What to Expect From Us
We strongly encourage you to apply! If we feel your skills, experience, and values match, we’ll reach out about meeting with the team.
During the interview stage, you can expect:
- An introductory interview with the hiring manager (50 minutes); if there’s a match we’ll schedule an interview loop with the team.
- A technical screen, either as a live interview via Karat or as a take-home assessment
- An Interview loop with 3-4 interviews (1 hour each) with the team members you will be potentially working with
All interviews will be remote via Google Meets; we are happy to make accommodations you might need to feel comfortable and set up for success in our process.
WellSaid Labs is honored to be an equal opportunity workplace. We realize that by bringing together teams rich in diverse thoughts and experiences, our people, company, and customers are free to flourish. We are committed to providing equal employment opportunities regardless of race, color, national origin, religion, creed, genetic information, sex (including pregnancy, sexual orientation or gender identity), age, marital status, disability, military or veteran status