About the roleYour mission is to turn Neon's raw consumer audio streams into the cleanest, most reliable training data on the market, and to build the commercial and operational engine that gets it into the hands of the world's leading AI labs.
As a Data Ops Lead, you'll own the end-to-end journey that takes raw recordings from our growing community of 500,000+ mobile users and delivers production-ready datasets to frontier labs. In practice, that means three things above all:
- Structuring and managing the data deals that turn our recordings into revenue
- Holding every dataset to a quality bar that keeps buyers coming back
- Standing up human transcription, annotation and other operations, largely overseas, that make it all possible
You'll work directly with our CEO on commercial priorities and help shape each deal, interface with buyer-side engineering and research teams at frontier labs to translate their exact specifications into deliverable dataset plans, and partner with internal engineering and external vendors to make sure the pipeline supports what we've sold. This is a foundational role: the datasets and processes you build are the product we sell.
You have...- Authorization to work in the US.
- 5+ years of experience building and scaling data pipelines for AI/ML applications, with significant time spent on audio, speech, or multimodal data.
- A track record of structuring and delivering against data or dataset agreements with external partners: taking their requirements, turning them into clear specifications, and owning delivery end to end.
- Experience building and managing overseas or outsourced teams for data tagging, annotation, and QA, with a track record of maintaining quality and throughput across time zones.
- Deep ownership of data quality: designing QA processes, defining acceptance criteria, and catching problems before a customer ever sees them.
- Enough technical fluency to be credible on both sides of a deal. You understand digital audio fundamentals (sample rates, VAD, multichannel formats), can reason about how pipelines are built, and know what "good" looks like, even if you're not writing every line of code yourself.
- A "Founder's Mentality." You're comfortable building from zero and making high-stakes calls with incomplete information.
Bonus points- A background working with audio data in some capacity.
- Direct experience with training data for TTS, ASR, speaker ID, or full-duplex conversational models.
- Familiarity with the modern audio stack (Librosa, FFmpeg, SoX, torchaudio) and cloud data infrastructure (S3, Redshift, BigQuery, or equivalent).
- An understanding of how high-quality, speaker-separated audio gets captured (for example, via WebRTC-based recording tools).
- Experience with active learning loops, human-in-the-loop QA systems, or corpus stratification for balanced dataset design.
- Prior experience leading a data or infrastructure team, including hiring and mentoring engineers.
